Research Article An Efficient Stock Recommendation Model...

Research ArticleAn Efficient Stock Recommendation Model Based onBig Order Net Inflow

Yang Yujun123 Li Jianping1 and Yang Yimei2

1School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu 611731 China2Department of Computer Science and Technology Huaihua University Huaihua 418008 China3Hunan Provincial Key Laboratory of Ecological Agriculture Intelligent Control Technology Huaihua 418008 China

Correspondence should be addressed to Yang Yimei yymhhtceducn

Received 14 August 2015 Revised 10 December 2015 Accepted 15 December 2015

Academic Editor David Bigaud

Copyright copy 2016 Yang Yujun et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

In general the stock trend is mainly driven by the big order transactions Believing that the stock rise with a large volume is closelyassociated with the big order net inflow we propose an efficient stock recommendation model based on big order net inflow in thepaper In order to compute the big order net inflow of stock we use the MG1 queue system to measure all tick-by-tick transactiondata Based on an indicator of the big order net inflow of stock we select some stocks with the higher value of the net inflow toconstitute the prerecommended stock set for the target investor user In order to recommend some stocks with which this style isfamiliar them to the target users we divide lots of investors into several categories using fuzzy clustering method and we shoulddo our best to choose stocks from the stock set once operated by those investors who are in the same category with the target userThe experiment results show that the recommended stocks have better gains during the several days after the recommended stockday and the proposed model can provide reliable investment guidance for the target investors and let them get more stock returns

1 Introduction

In the area of stock recommendation method research [1]most of the research mainly focuses on the two areas stockrecommendation methods based on stock comment [2] andprice forecasting [3] The former is easy to understand andmaster for investors However in such a complicated stockmarket the investors do not know which one to believeamong the lots of stock comments with dubious authenticityand every choice has a great risk for themThe latter method[4] is difficulty for investors to understand and master sincethe application of the latter method is relatively complex andinvolves a lot of profoundmathematical knowledge [5] Giventhis situation many scholars have done a lot of research onthe stock recommendation [6]

Currently the stock recommendation based on priceforecasting relies [7] mainly on mathematical and statisticalmethods [8] time series model [9 10] and machine learningmodel [11] Sonsino and Shavit [12] have researched a stock

prediction and selection method based on unidentified his-torical data M-Y Chen and B-T Chen [13] have proposeda stock price forecasting method based on the hybrid fuzzytime series and granular computing Xin et al [14] havegiven a strategy for filtering out users with similar demandcharacteristics by using collaborative recommendation algo-rithm with fuzzy clustering method which shows excellentrecommendation effect

In present financial field how to integrate multipletechnologies [15] such as data mining machine learning andherd psychology [11] and other nontraditional technologiesinto stock recommendation has become a hot topic Fewpapers use money net inflow as stocks recommendationtechniques Given this situation we proposed an efficientstock recommendation model based on big order net inflowin the paper At first we divide lots of users into severalcategories utilizing collaborative filtering algorithm based onuser fuzzy clustering [16]We get some stocks from the stocksonce operated by those users in same category and form a

Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2016 Article ID 5725143 15 pageshttpdxdoiorg10115520165725143

2 Mathematical Problems in Engineering

prerecommended stock set Then we use a method based onMG1 [17] to compute the net inflow amount of big orderfor every stock in the prerecommended stock set From theprerecommended stock set descending ordered by the valueof big order net inflow we choose some stocks with highestvalue in front of the set as the last recommendation stock setfor the target user

In general we believe that the stock rise with a largetrading volume is closely related to the purchase stock ofbig order In order to analyze the big order net flow ofstock we need to observe the stock trading volume andturnover The money net flow and the money flow [18] aredifferent in concept The big order refers to the amount ofeach transaction over one million yuan or the volume of eachtransaction over fifty thousand in a single transaction So thebig order net inflow refers to the amount of money of bigorder buy or sell of the same stocks within a day Most ofthe time the money flow is bigger than zero and the moneynet flow is less than zero In individual cases the money netflow is bigger than zero and the big order net flow is lessthan zero Under this situation we proposed an efficient stockrecommendation model based on big order net inflow Thenew recommendation model can measure the capital andthe pulsation of the stock markets and consider investorspreferences and behavior characteristics it can improve theexisting deficiencies of some current stock recommendationIn addition the new recommendationmodel can analyze andfilter the stock with less returns in the future and improvethe investment gains of investors The experiment resultsshow that the recommended stocks have better gains duringthe several days after the recommended stock day and theproposedmodel can provide reliable investment guidance forthe target investors and let them getmore investment returns

The rest of this paper is organized as follows Section 2briefly reviews the definitions and theorems of fuzzy cluster-ing and the framework of the collaborative filtering algorithmbased on user fuzzy clustering Section 3 demonstrates themethod for computing big order net inflow and the frame-work of the proposed model Section 4 presents the simu-lation experiment and empirical analyses of the proposedmodel finally some conclusions are given and some futureworks are pointed out in Section 5

2 Theoretical and Modeling Framework

The concept of fuzzy set [16] in fuzzy cluster is put forward byZadeh in 1965 Gath and Bar-On earlier applied that theoryto compute the scoring of poly graphic sleep recordings intheir study [19] In this section we will briefly review thedefinitions and theoretic of fuzzy cluster

21 Fuzzy Clustering Theory

Definition 1 Defined set 119877 = (119903119894119895)119898times119899

a matrix with 119898 rowsand 119899 columns if 0 le 119903

119894119895le 1 then 119877 is called fuzzy matrix

119877 When 119903119894119895is only 0 or 1 said 119877 is a Boolean matrix When

elements 119903119894119894of the diagonal are all 1 in fuzzy matrix 119877 =

(119903119894119895)119899times119899

said 119877 is reflexive fuzzy matrix

Definition 2 If119860 is 119899-order square defined1198602= 119860 ∘1198601198603 =

1198602∘ 119860 and 119860

119896= 119860119896minus1

∘ 119860


and called 119877119879

=

(119903119879

119894119895)119898times119899

is the transposed matrix of 119877 where 119903119879119894119895= 119903119894119895


for any 120582 isin [0 1] andsaid 119860

120582= (119903(120582)

119894119895)119898times119899

is the 120582-cut matrix of the fuzzy matrixR When 119903

119894119895ge 120582 119903(120582)

119894119895= 1 and 119903

119894119895lt 120582 119903(120582)

119894119895= 0 said 119860

120582is a

Boolean matrix and said 120582 is a confidence level parameter orcut level parameter

Definition 5 Suppose two limited discourse domains 119883 =

1199091 1199092 119909

119898 and 119884 = 119910

1 1199102 119910

119899 then 119883 to 119884 fuzzy

relation 119877 is an119898times119899 order fuzzy matrix and set 119877 = (119903119894119895)119898times119899

where 119903

119894119895= 119877(119909

119894 119910119895) isin [0 1] represents (119909

119894 119910119895) relevance on

fuzzy relation 119877

Theorem 6 If 119877 is fuzzy similar matrix then for any naturalnumber 119896 119877119896 is fuzzy similarity matrix

Theorem 7 If 119877 is an 119899 order fuzzy similar matrix then thereis minimum natural number 119896 (119896 le 119899) for all natural number119898 (119898 ge 119896) constant 119877119898 = 119877

119896 119877119896 namely fuzzy equivalentmatrix (1198772119896 = 119877

119896) At this point said119877119896 is the transitive closure

of 119877 denoted by 119905(119877) = 119877119896

Method 8 If 119877 is fuzzy similar matrix then the followingmethod will solve the transitive closure 119905(119877) 119877 rarr 119877

2rarr

1198774rarr 1198778rarr 11987716

rarr sdot sdot sdot rarr 1198772119896

and the said method is thesquare self-synthesis method (S3M)

22 Fuzzy Clustering Analysis The fuzzy clustering analysisis an analyzing clustering and classification method by estab-lishing fuzzy similar relationship of objective things basedon the objective characteristics the degree of closeness andsimilarity between objective things

In Figure 1 the fuzzy clustering processing model can bedivided into four stages namely the data preprocessing thedata standardization the constructing fuzzy similar matrix(FSM) and the clustering and analysis

221 Data Preprocessing In China it is well known thatindividual investors buy or sell stocks only through a secu-rities company not directly from the stock exchange Inorder to trade stocks individual investors firstly have toregister as a member in a securities company Accordingto the National Security Act investors must have a testin the risk tolerance when investors are registering Thesecurities company gives them a risk test paper with fifteenquestions or more Each question has four or five options forinvestors to choose In order to store and process we usean integer value instead of the chosen answer of investorsin the user risk database In this paper we choose twelveanswers among all answers to study There are at least 1million records about risk information of investors in theuser risk database of any securities company because there

Mathematical Problems in Engineering 3

Table 1 The surveying contents of investor

119909119894119895

Denoted contents Range1199091198941

The age range of investor [1 5]1199091198942

The funds amount of investor that can be used to invest [1 5]1199091198943

The investment purposes of investor [1 4]1199091198944

The duration of your general investment [1 4]1199091198945

Themaximum loss that you can accept within one year [1 5]1199091198946

The selection of four portfolio average returns with the best and worst of earnings in the next three years [1 4]1199091198947

The annual income change of your family in next five years [1 4]1199091198948

The ratio of domestic consumption expenditure share of total income monthly in your family [1 5]1199091198949

The highest risk products that you have invested before [1 4]11990911989410

The annual income of your family [1 5]11990911989411

The number of years since you invest [1 4]11990911989412

The type of your work [1 5]

Clustering

(10) Compute transitive closure

(11) Set clustering threshold value

(12) Get clustering category

Constructing FSM

(7) Compute Euclidean distance

(8) Set parameter value

(9) Construct fuzzy similar matrix

Standardization

(4) Get the minimum data

(5) Get the maximum data

(6) Execute range transformation

Preprocessing

(1) Collect and analyze the data

(2) Define the discourse domain

(3) Construct the raw data matrix X

Figure 1 The structure of fuzzy clustering processing model

are more than 200 million stocks investors in China So weassume that an investor risk tolerance surveying databaseexists which includes 119899 investors risk tolerance surveyingdata The characteristics of these data which include thefollowing twelve contents reflect the investment style and risktolerance of the investors We define the set of investor risktolerance surveying data as 119883 = 119909

1 1199092 119909

119899 where 119909

119894

denotes the surveying data of the ith investor in the set 119883

which is constituted by the following contents vector in a fixedorder 119909

119894= 1199091198941 1199091198942 1199091198943 119909

119894119895 119909

11989412 where 119895 belongs to

[1 12] Table 1 shows the detailed contents of the ith investorin the investor risk tolerance surveying database

222 Data Standardization We firstly compute all columnsdata of raw matrix 119883 get the minimum data and themaximum data of each column and then compress eachdata of matrix 119883 to [0 1] using the following transformationformula After the above standardizing process we can getstandardized matrix 119884 with standardizing data

119884 = 119910119894119895 1 le 119894 le 119899 1 le 119895 le 12

119910119894119895=

119909119894119895minusmin

1le119894le119899119909119894119895

max1le119894le119899

119909119894119895 minusmin

1le119894le119899119909119894119895

1 le 119895 le 12

119884 =

[[[[[[

[

11991011

11991012

sdot sdot sdot 119910112

11991021

11991022


1199101198991

1199101198992


]]]]]]

]

(1)

223 Constructing Fuzzy Similar Matrix After obtainingstandardized matrix 119884 with standardizing data we use thedirect Euclidean distance method as the similarity coeffi-cient method to determine the similarity coefficient amonginvestors and construct the fuzzy similar matrix Consider

119903119894119895= 1 minus 119888 times 119889 (119909

119894 119909119895) (2)

where 119888 is a suitable choice of parameters so as to 0 le 119903119894119895le 1

and the 119889(119909119894 119909119895) represents the distance between 119909

119894and 119909

119895as

follows

119889 (119909119894 119909119895) = radic

12

sum

119896=1

(119909119894119896minus 119909119895119896)2

119888 =1

1 + 1198881015840

(3)


where 1198881015840 belongs to [0infin) and 119888 belongs to (0 1] We can

choose any value for 1198881015840 If the 1198881015840 value is too high parameter119888 will be less which will lead to increasing the accuracy ofcomputing Therefore we chose an appropriate value for 1198881015840In fact we can choose zero value for 1198881015840 which does not affectthe experiment results 119888 is equal to 1

Thus we can get fuzzy similar matrix 119877 of 119899 by 119899 betweeninvestors

119877 =

[[[[[[

[

11990311

11990312


11990321

11990322


1199031198991

1199031198992

sdot sdot sdot 119903119899119899

]]]]]]

]

(4)

224 Clustering and Analysis Since fuzzy similar matrix 119877

is a 119899-dimensional square matrix we can carry out transitiveclosure 119905(119877) using the square self-synthesis method

119905 (119877) 119877 997888rarr 1198772997888rarr 119877

4997888rarr 119877

8997888rarr 119877

16997888rarr sdot sdot sdot

997888rarr 1198772119896

(5)

where 119896 le [log1198992]

According to the actual situation we have to choosean appropriate 120582 value between 0 and 1 for 120582-cut matrix119905(119877)120582of transitive closure 119905(119877) of fuzzy matrix 119877 Having

a classification of 119905(119877) based on 120582 we can get equivalenceclassification matrix 119905(119877)

120582under the given 120582 value

Thus we will get clustering category for investors Oneday a new investor who becomes the target recommendationinvestor of stock recommendation system will be added tothe investor risk tolerance surveying database In order todetermine which category the new investor belongs to weutilize the fuzzy clustering method to subdivide the databaseinto several groups based on the above 120582-cut value

23 Collaborative Filtering Algorithm

231 Nearest Neighbors Choice Currently the collaborativefiltering algorithm is themost successful personalized recom-mendation algorithm It can be classified into two categoriesone is item-based collaborative filtering algorithm [20] theother is user-based collaborative filtering algorithm Theformerwas first put forward by Sarwar et al [21] that whenwecalculate the user similarity first we calculate the similaritybetween items to select the most similar items and thenpredict the ratingThe latterwas first proposed byGoldberg etal [22] which is according to the rating of target userrsquos nearestneighbors to predict the target item ratingThis algorithm canbe computed offline shorten the time of online calculationand increase speed of the online recommendation

In order to improve the accuracy of the nearest neighborinvestorrsquos choice we will use the user-based collaborativefiltering algorithm to compute the degree of similarity inbehavior between other investors and the target investorin the same fuzzy cluster Then according to the stocks ofthe nearest investors and target investor we can generate

optimized stock list from the stock set calculated by fuzzycluster algorithm and choose the top-119896 stock to recommendthe target investor

At first we need to calculate the degree of similaritybetween investors based on their risk tolerance surveyingdata At present there are several methods to calculate thesimilarity between investors such as cosine-based similaritythe adjusted-cosine similarity and Person correlation-basedsimilarity [23] According to the degree of similarity betweenthe target investor and other investors we can generateneighbor set 119880 = 119880

1 1198802 1198803 119880

119899 for the target investor

119906 Then we choose the top-119896 investors as neighbor investorfor the target investor based value of sim(119906 119880

119894) ordered by

descending

232 Constructing Stock Cluster Set In order to recommendsomemore accurate stocks to the target investor we subdividethe stocks into several categories as follows based on thefuzzy clustering method in Section 22 in this paper Thenwe can utilize fuzzy clustering analysis method to constructstock cluster set As we know there are dozens of attributesin each stock We choose six important attributes of stockto construct a stock attributes database We think the sixattributes of stock are the most important attributes for stockclustering The six attributes are the daily average gainsthe daily average amplitude the days of price rise the netprofit in last year the daily net amount of big order andthe days with net amount of big buying order According tothe six attributes of stock and the data format like Table 1in this paper we construct the stock attributes databaseThen we cluster the stocks in the stock attributes databaseand subdivide the stocks into several categories We caneffectively distinguish different stocks between poor stocksand good stocks Such clustering results can reflect thosestocks operational characteristics and be able to providemore accurate and effective recommendation informationfor target investor In generally the stocks in same clusterhave similar trading dynamic characteristic If we have arecommended stock for target investors we will try to findsome other stocks in its stock cluster set This can improvethe accuracy of recommended stocks and reduce the difficultyof the search for other recommended stocks In order toreduce the computational complexity of clustering the stockswe have to guarantee that the number of cluster is less than10 The number of cluster will vary with the changes ofrecommended stocks Generally the number of cluster variedbetween five and seven in experimentation

233 Generating Recommended Stocks List At first we getthe stocks list of the target investor and then we get the scoreof each stock in stocks list of target investor and the score ofsame stock like in stocks list of neighbor investors If somestocks are not in the stocks list of neighbor investors we canforecast their scores using the following formula

119891 (119906 119895) =

radicsum119870

119896=1sim (119894 119896)

2lowast (119878119896minus 119878119896)2

radicsum119870

119896=1sim (119894 119896)

2

+ 119878119906 (6)


where 119878119906represents the target investor 119906rsquos average scores for

all stocks 119878119896represents the 119896th neighbor investor scores of

the target investor 119906 119878119896represents the average scores of 119896

neighbors investor for all stocks and 119891(119906 119895) represents the119895th stocks forecast score of target investor 119906 sim(119894 119896) is amodified similarity formula based on the cosine Similarityit is defined as follows

sim (119894 119896) =

sum119898

V=1 (119877V119894 minus 119877119894) (119877V119896 minus 119877

119896)

radicsum119898

V=1 (119877V119894 minus 119877119894)2radicsum119898

V=1 (119877V119896 minus 119877119896)2

(7)

where 119877119894is the 119894th stock average scores for all investors 119877

119896is

the 119896th stock average scores for all investors 119877V119896 representsthe 119896th neighbor investor scores for the Vth stock and 119877V119894represents the 119894th neighbor investor scores for the Vth stock

Now we give a simple example for the process of generat-ing recommendation stock list Assuming that there are twostocks in the stocks list of the target investor such as stock 119860

and stock 119861 here we write them as a set 119860 119861 Then we getthe stocks set of four or more neighbors such as set 119861 119862119863119861 119862 119864 119862 119865 and 119861 119862 and we can select the rated top-119896stocks as the target investorrsquos extend stocks list from the abovefive stock set When we set 119896 as 2 we can get the top-2 stocksset 119861 119862

Secondly in order to get themore large stock extend set ofthe target investor such as stock set 119861 119862119883 119884 119885 we need torevise the 119896 value If we recommend to the target investor onlyfive stocks we will stop revising the 119896 value when the numberof stock in the above stock set is greater than or equal to five

Finally we calculate net inflow amount of big orderfor each stock of the above stock extend set 119861 119862119883 119884 119885

recently then we sort the five stocks in descending orderaccording to the net inflow amount of big order and selecttop-119899 stocks recommend to the target investor If we set 119899as three we get the recommend stock list 119861 119883 119884 Becausethe process of judging money inflow is more complex wepropose a new method based on MG1 to compute the netinflow amount of big order and we can forecast the directionof stock price movements in the future

24 MG1 Queue System The MG1 queue system [24]has been extensively studied for the last three decades [25]According to the circumstances of the securities transactionhere we review the single server queue system which behaveslike the usual MG1 queue when the server is working Weassume that the server goes through cycles of idle and busyperiodsThe idle periods include two cases one is when thereis no work to do and the other is when there is work to do butthe server is on vacationThe busy periods are the times whenthe server is actually working on the customers of primarycustomers [26] During the busy periods of the simple MG1queue system we assume that the customers arrival timefollows Poisson distribution with parameter 120582 Let 120582 be thecustomer arrival rate and let 119863 be the distribution of theservice times of customers arriving during busy periods thenthe customers arrival time has a general distribution function119861 Let 119878

119894be the epoch at the end of the 119894th busy period

and let 119879119894be the epoch beginning the 119894th busy period Then

0 = 119879119894

lt 119878119894

lt 119879119894+1

lt 119878119894+1

lt sdot sdot sdot lt 119879119899

lt 119878119899 We

assume that the arrival process and the service times of thecustomers arriving in the interval (119879

119894 119878119894) are independent of

those arriving in (119879119895 119878119895) for 119894 = 119895 Let 119908(119905) be the work in the

queue system at time 119905 Let 1198821119894

= 119908(119878119894) and 119882

2119894= 119908(119879

119894+1)

where 119894 gt 0 Then1198821119894is the work in the queue system at the

end of the 119894th busy period and 1198822119894is the work in the queue

system immediately after the beginning of the (119894 + 1)st busyperiod

Let 119897(119896) be the 119896th workload step in all workload processLet 119871(119896) = min119898 sum

119898

119894=1(119878119894minus 119879119894) ge 119896 and 119864(119896) =

sum119871(119896)minus1

119894=1(119879119894+1

minus 119878119894)

If 119896 lt sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896) = 119908(119896 + 119864(119896))

If 119896 = sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896

+) = 119908(119896 + 119864(119896) + 119879

119871(119896)+1lt

119878119871(119896)

) and 119897(119896minus) = 119908(119896 + 119864(119896))

Clearly if 119896 = sum119871(119896)

119894=1(119878119894minus119879119894) then 119897(119896

+) minus 119897(119896

minus) = 119882

2119871(119896)minus

1198821119871(119896)

The 119897(119896) process behaves like the work in a simpleMG1 queue

3 Proposed Method

This section consists of two subsections Section 31 describesthe method based on MG1 to compute the net inflowamount of big order and Section 32 describes the proposedmodel and method for stock recommendation

31 Compute the Net Inflow Method

311 Funds Flow Theory The flow of funds is stock move-ment direction actively chosen by the funds in the stockmarket From the amount of perspective to analyze the flowof funds namely observation volume and turnover tradingvolume and turnover in the actual operation is directionalto buy or sell For both the stock market trend analysis andthe operation on individual stocks the determination of thefunds flow plays a vital role and the process of the funds flowis more complicated not easy to grasp The funds flow canhelp investors see what others are doing in the end throughthe index (price) change fog For example the index (price)of a stock rise up to a point may be driven by 10 millionfunds or a billion of funds both of which have a completelydifferent significance for investors In general funds flowand the trend of stock index change are very similar butin the following two cases funds flow measure has obvioussignificance One is that the dayrsquos flow of funds and stockindex change opposite For example the stock overall index isdown throughout the day but funds flow shows a positive netinflow of funds throughout the day The other is that there isvery big opposite between the funds flow and the stock indexchange For example the stock index rises highly throughoutthe day but the actual net inflows are small or even negativewhich is called net outflows When the funds flows and thestock index change is opposite the funds flow can reflectstock actual movement direction more than the stock indexchange in the future


312 Funds Flow Concepts In order to more clearly describethe proposed method we review some concept about fundsflow as follows

Definition 9 Funds inflow refers to the amount of activebuying It is active buying transactions where the buyeractively buys stock with price equal to or higher than the firstselling price

Definition 10 Funds outflow refers to the amount of activeselling It is active selling transactionswhere the seller activelysells stock with price equal to or less than the first buyingprice

Definition 11 Funds net inflow refers to the amount of fundsinflow minus funds outflow If it is positive the probabilityof stock price rise is higher than fall and vice versa if the netinflow is negative then the probability of stock price fall ishigher than rise

Definition 12 Big order refers to the buying or selling orderwith big amount According to the amount of the order wedivide it into four categories small order general order bigorder and very big order (king order) Using the number ofthe order which measures it we think that the number of bigorder is more than 5 million shares or more than 20 millionyuan and the number of king order is more than 20 millionshares or more than 100 million yuan

Definition 13 Big order net amount refers to the differencebetween the amount of big buying order and the amount ofbig selling order If it is positive we call it the big order netinflow If not we call it the big order net outflow

Definition 14 Tick-by-tick data refer to the single transactionduring transactions It reflects the true circumstances of thetransaction process and is proprietary data of the Level-2

313 Method Based onMG1 We review the classical MG1queue system which has the following four characteristics[27]

(1) The characteristic of the arrival process follows Pois-son distribution with arrival rate parameter 120582 Mindicates a Poisson process without memory

(2) Probability 119875 of the service time follows generalrandom distribution Let 119905

119894be the service time for the

119894th customer that is an independent and randomvalueand has general distribution function 119861(119905)

119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)

Themean value and variance of service time are givenby

119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)

(3) The number of server desk is one the arrival time andservice time are independent of each other

(4) The system allows an infinite captain for the length ofcustomers the queue discipline is first comefirst serve(FCFS)

Assuming that interarrival time follows Poisson distribu-tion with parameter 120582 in 119860(119905) 119905 ge 0 let119860(119905) be the numberof arrival customers in [0 119905] time let 119873(119905) be the number ofcustomers in queue system at time 119905 let 119883

119899be the number

of customers after the 119899th customer departure instant let 119889119899

be the departure time of the 119899th customer and let 119886119899be the

arrival time of the 119899th customer If the number of customersis greater than zero at time 119889

119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)

At this time if119883119899= 0 then

119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)

From (11) and (12) the following can be obtained

119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)

Transition probability matrix 119875 of Markov Chain119883119899is

119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot

0 0 0 0 sdot sdot sdot

))))))

)

(14)

where chain119883119899has Markov property and 119901

119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)

32 The Framework of the Proposed Model In Figure 2 theprocess and data flow framework of the proposed modelcan be divided into two stages namely the user clusteringstage and stock recommend stage Generally investors whohave similar characteristics have similar investment interestAccording to Figure 2 in order to obtain accurately thesimilar stocks for target users we have to process userclustering and obtain some similar users The selection ofthe clustering threshold value affects the number and sizeof the user category and then affects the accuracy of thestocks set which is selected from the same category userstocks so it is critical to select the value of user clusteringthreshold In the stock recommend stage the computing ofthe big orders net for lots of stocks is themost important partThe stock trend is mainly driven by big order transactions


User clustering

Stock recommend

(4) Get user clustering category

(5) Get target user category

(6) Construct user stock set

(1) Collect investor user data

(2) User data standardization

(3) Construct user FSM

(7) Compute big order net funds

(8) Get recommend stocks set

(a)

Compute inflow big order

Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories

Target user stocks list

Neighbor userlist

Target user stocks extend list

Recommended stocks list

Finding

Neighbor user stocks list

Stock DB

Stock categories

Clustering

(b)

Figure 2 (a) The process framework of the proposed model (b) The data flow framework of the proposed model

It is generally believed that stocks rise with a large volumeis closely associated with big orders net amount to buyso the stocks are generally rising in price under the trenddriven by big orders net inflow which is called big ordersnet buying In contrast the stocks are generally falling inprice under the trend driven by big orders net outflow whichis called big orders net selling The big orders net amountincludes the big orders net buying and the big orders netsellingThe traditionalmodel uses the funds inflow and fundsoutflow to predict stock trend that is unsuitable for somenew special situation For example one day the funds inflowof a stock is far greater than zero but the big orders netinflow of the stock is far smaller than zero If it is that casewe can predict that this stock will go to falling trend over aperiod of time after that day After observing many stockswe found that it is indeed the case Our proposed modelcan solve this problem by using big orders net amount thatcan avoid or reduce some forecast errors and can improveaccuracy of recommend stocks trend in the future Thenaccording to the indicated results of big orders net inflow weselect some optimal stocks recommend to target users to buyCorrespondingly according to the indicated results of bigorders net outflowwe select some optimal stocks recommendto target users to sell

4 Simulation Experiment

In this section we study and compare the performanceof the proposed model In general during the Shanghai

247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965

Figure 3 The research interval price trends graph of the ShanghaiComposite Index (CSI)

and Shenzhen Composite Index (CSI) rise the accuracy ofthe recommendation algorithm is higher but during theCSI fall the accuracy of recommendation algorithm is verylow or even completely incorrect Thus in the course offalling the experimental results can test a recommendationalgorithm In order to make the experiment results moreobjective and realistic we use stock return to test whether therecommended stock has a good return from 10 to 30 days afterthat For target investor the higher the yields the investor getsthe better the effect of the recommendation model

41 Data Selection In order to examine whether the pro-posed model has made improvement in prediction accuracywe select data at four different periods of the real stockmarketin China as the experiment data The four periods includebottom period (20121128ndash20121204 see A in Figure 3)middle period (20121217ndash20121221 see B in Figure 3)top period (20130204ndash20130208 see C in Figure 3) of


Table 2 The users clustering results

Category number Users numbers Percentage ()1 912 35632 138 5393 576 22504 482 18835 314 12276 48 1887 90 352

a wave with rising trend and middle period (20130617ndash20130621 see D in Figure 3) of a wave with falling trendThe experimental data include data of three parts that arethe investor user data the CSI data and the tick-by-ticktransaction data of stock As the userrsquos data are related to theuserrsquos personal privacy we use the data which are removedfrom the userrsquos privacy as the experimental dataThe CSI andstocks data are stock market free data but the tick-by-ticktransaction data of stock are the Level-2 data and chargingdata

42 Data Processing In order to reduce the amount ofcomputation and be suitable for parallel computing werandomly select 2560 users from the investor user database tofuzzy clustering and divide those users into several categoriesaccording to the threshold of clustering By the nature ofthe clustering the value of the threshold can control thenumber of the user categoriesThe value of the threshold andthe number of the user categories are inversely related Thethreshold can be in the range from 0 to 1 If the threshold isset too high we will get very few user categories Contrarilyif the threshold is set too small we will get much moreuser categories For example if the threshold is set to 1we will get 2560 user categories Clearly if the threshold isset to 0 we will get one user categories The complexity ofclustering the users increases exponentially with the numberof the user categories In order to reduce the computationalcomplexity of clustering the users we have to choose anappropriate threshold For that 2560 users data we madeseveral experiments to cluster that users data by constantlyadjusting the value of the threshold We can get seven usercategories and get better distribution of user in Table 2 whilethe threshold was set to 06268 Certainly it is allowed thatthe threshold is set higher or lower than 06268 but thedistribution of user will become worseTherefore in order toachieve the purpose of the clustering users we have to choosean appropriate clustering threshold After making severalexperiments we set the clustering threshold 120582 = 06268 inthis paper and then divide 2560 users into seven categories inTable 2

After the users clustering we use formula (7) based on thecosine similarity to calculate the similarity between the targetuser and the other usersWe can find out the nearest neighborusers of the target user according to the similarity value In thesimilarity calculation here there are two cases (1) the targetuser belongs to the clustered users (2) the target user does not

Table 3 The nearest neighbors of the target user

No Similarity value Users number1 0982 3802 0957 3593 0926 24784 0903 19115 0871 16536 0819 4067 0735 141

belong to the clustered users For case (1) the target user andhis most nearest neighbors must belong to the same categoryof the clustered users we only need to find out the 119896 mostnearest neighbors in the same category according to the valueof the fuzzy similarity matrix For case (2) we calculate thesimilarity value between the target user and the log119899

2clustered

users at most based on the Binary Search Algorithm [28]where 119899 is the number of clustered users In order to reducethe amount of calculation of the recommendation actionsafter that we add the target user into the same category withthe most clustered user In order to make the experimentresults more objective and realistic we choose an investoruser that does not belong to the clustered users as the studytarget in the experiment and the result of the nearest sevenneighbors of the target user in Table 3

With the complement classified for the target user wecan get the 119896 clustered users and get the stock set from thoseusers Here we call stock set as a prerecommended stock setWe use a method based on MG1 to compute the net inflowamount of big order for every stock in the prerecommendedstock set From the prerecommended stock set descendingordered by the value of big order net inflow we choose the119896 stocks with highest value in front of the stock set as the lastrecommendation stock set for the target user Due to limitedpaper space we selected five stocks in each period of Figure 3as the research stock in Tables 4 5 6 and 7 If the value of anystock in the stock set is smaller than 0 we have to set that thevalue of 119896 is bigger and repeat the previous calculation stepsIf the prerecommended stock set contains the all stock of theChina Stock Market and any stock value of the big order netinflow is smaller than 0 we do not recommend any stock tothe target user and recommend the target user to keep a waitstate with holding money

We do some research on some stocks that are recom-mended by the proposed model to analyze the maximumpossible stocks return during the four different day intervalsthat is 10 days 30 days 90 days and 1 year since that daywhenwe recommend those stocks to the target user By observingthe experiment results we found that those recommendedstocks have better gains in Table 8 especially those stockswhose big order net inflow is higher proportion of tradingvolume of those days than the others In addition the stockprice is the highest stock price without ex-rights or ex-dividend at the last trading day of the four corresponding dayintervals


Table 4 The recommended five stocks in the A period

Stock name (code) Datemean Buy Sell Net amount (104 share) Percent () Net money (104 yuan)

Jianghuai Automobile (600418)

2012-11-28 6587 16200 minus9611 minus1909 minus481402012-11-29 7798 2248 5550 852 283602012-11-30 47620 21660 25950 2137 1359302012-12-03 96530 45010 51520 1893 2801702012-12-04 109880 50580 59300 2459 328590

Mean 53683 27140 26542 1086 144982

Janus Precision (300083)

2012-11-28 000 000 000 000 0002012-11-29 649 000 649 1722 68472012-11-30 000 000 000 000 0002012-12-03 335 000 335 1275 35172012-12-04 326 400 minus074 minus154 minus734

Mean 262 080 182 569 1926

Great Wall Motors (601633)

2012-11-28 5862 5176 686 121 123002012-11-29 9353 10920 minus1570 minus420 minus279102012-11-30 12540 9961 2577 572 471402012-12-03 25920 21680 4235 760 780502012-12-04 30140 23600 6543 1303 119620

Mean 16763 14267 2494 467 45840

Huasun Group (000790)

2012-11-28 7127 4275 2852 824 179002012-11-29 000 1034 minus1034 minus472 minus62212012-11-30 662 000 662 284 40482012-12-03 2160 2388 minus228 minus081 minus14042012-12-04 2426 1311 1115 454 6376

Mean 2475 1802 673 202 4140

Faw Car (000800)

2012-11-28 979 4120 minus3141 minus1068 minus184702012-11-29 2303 2487 minus184 minus164 minus18102012-11-30 3049 4012 minus964 minus257 minus56062012-12-03 24180 24550 minus373 minus043 minus16622012-12-04 31490 14130 17360 1721 107510

Mean 12400 9860 2540 038 15992

43 Experimental Results Analysis In this section we brieflyanalyze and explain the experimental results After selectingan investor user that does not belong to the clustered usersas the study target user in the experiment we calculate thesimilarity value between the target user and the clusteredusers using formula (7) based on the cosine similarityBecause the nearest neighbors of the target user all belongto category 3 of clustered users the target user belongs tocategory 3 according to the nature of the clustering that thetarget user and his nearest neighbors belong to the samecluster We select the nearest seven neighbors of the targetuser to show you in Table 3We use amethod based onMG1to compute the net inflow amount of big order for everystock in the prerecommended stock set selected from thoseusers We select five stocks with higher value of big order netinflow during the period between 20121128 and 20121204as the recommended stocks for the target user in Table 4For simplicity we sort the stocks using big order net inflowdaily mean ratio from large to small and select five stocks torecommend the target user

Table 4 shows the daily big order net flow ratio andmoney for the five recommended stocks In order to ensurethe data neat and objective we calculate the daily big ordernet flow for every stock at continuous five trading days andcompute the average of them Combining the big order netmoney inflow we use the big order net inflow ratio as firstevaluation criterion for the stocks From the data in Table 3we find that the number of days of big order netmoney inflowis generally greater than the number of days of big order netmoney outflow Although there are four days of big ordernet money outflow in the stock of Faw Car the big ordernet money inflow of the fifth day is greater than the sum ofbig order net money outflow of the previous four days Forthat reason we believe that the stock will rise in the futureGenerally the big order net inflow only affects the short-termprice of stock and shows that some investors are upbeat aboutthe stock in the near future The short-term price of stockmay be consistent with how much the big order net moneyinflow is but the long-term price of stock may not be fullyconsistent with it As we all know the long-termprice of stock


Table 5 The recommended five stocks in the B period


Dahu Aquaculture (600257)

2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144

Changan Automobile (000625)

2012-12-17 43220 30040 13180 711 720602012-12-18 42710 84050 minus41340 minus1784 minus2185602012-12-19 527220 232990 294230 3197 17000002012-12-20 223860 190690 33170 630 1963602012-12-21 429770 268930 160840 2243 1000000Mean 253356 161340 92016 999 549972

Xinning Modern Logistics (300013)

2012-12-17 24670 10170 14490 4329 1243502012-12-18 15390 15150 241 028 21332012-12-19 2056 2987 minus932 minus230 minus80772012-12-20 909 1362 minus453 minus096 minus40182012-12-21 1073 835 239 043 2355Mean 8820 6101 2717 815 23349

Robam Appliances (002508)

2012-12-17 14170 7990 6183 1693 1150702012-12-18 1243 2404 minus1161 minus903 minus212802012-12-19 1414 390 1024 1552 189802012-12-20 170 1435 minus1265 minus1301 minus228302012-12-21 7950 7640 309 108 6014Mean 4989 3972 1018 230 19191

Great Wall Motor (601633)

2012-12-17 13940 14770 minus827 minus162 minus133002012-12-18 12570 13620 minus1052 minus195 minus218802012-12-19 9763 5514 4250 1017 893702012-12-20 25370 17960 7404 1404 1608702012-12-21 14640 14220 421 106 9046Mean 15257 13217 2039 434 44821

may be consistent with companyrsquos development companyrsquosprofitability stock traderrsquos goal and skill and so on We cansee the above phenomenon from Table 8 Correspondinglywe can find that similar characteristics exist in Tables 5 6and 7

Table 8 shows the recommended stocks maximum pos-sible gains during the four different day intervals In orderto conveniently compare with the recommended stocks weput the corresponding data of Shanghai Composite Index onthe bottom of the data of the recommended stocks Due toa similar trend of CSI 800 and Shanghai Composite Indexwe chose Shanghai Composite Index as CSI in Table 8 If thehighest price of the stock or CSI is equivalent in two or morecontinuous periods this shows that the price of the stock orCSI is smaller than the previous price of that one and showsthat the stock has been in decline during the following periodFor example the highest price of the CSI of the ldquohighest pricewithin 90 daysrdquo is equal to the highest price of CSI of theldquohighest price within 1 yearrdquo they all are 244480 and we findthat the running range of CSI is the interval from C to D inFigure 3 From Table 8 we find that the maximum possible

gains of most of the recommended stocks are bigger than theCSI Without a doubt the target investor cannot get the samegains with the maximum possible gains because the time ofbuying or selling stock is uncertain for the target investor Itis easy to understand that the recommended stocks can getbetter gains between A period and C period in Figure 3 Butit is very difficult that the recommended stocks can get bettergains between C period and D period or after D period sincethe CSI has been going down or fluctuantThey fully illustratethe effectiveness of the proposed model

In order to show clearly the return of the recommendedstocks we choose the C and D period in Figure 3 For that isthe start period andmiddle period of the CSI fall or fluctuantthe general stocks will fall or fluctuate with the CSI If therecommended stocks can get better returns in the fallingperiod it clearly shows that the proposedmethod is excellentWe select two last stocks from the recommended five stocksin the C and D period Figures 4 and 5 show the returncomparing the two stocks with the CSI in the 20 weeks afterrecommending those stocks Because the CSI market wasclosed for a holiday Figure 4 shows only 18 trading weeks and


Table 6 The recommended five stocks in the C period


Irtouch Systems (300282)

2013-02-04 2781 2255 527 926 78492013-02-05 639 554 085 273 12772013-02-06 2482 2577 minus095 minus143 minus13602013-02-07 918 236 682 1688 103202013-02-08 1823 783 1040 1958 16100

Mean 1729 1281 448 940 6837

Transinfo Technology (002373)

2013-02-04 1578 1901 minus323 minus541 minus36052013-02-05 328 737 minus409 minus1308 minus47062013-02-06 1939 584 1355 2365 159302013-02-07 4776 3935 841 894 99272013-02-08 3128 2418 710 952 8524

Mean 2350 1915 435 472 5214

Microgate Technology (300319)


Mean 2001 1670 330 417 5848

Kingsun Optoelectronic (002638)


Mean 25578 23710 1868 352 19714

Victory Precision (002426)


Mean 3366 2378 988 347 6670

Number of trading week1 18171615141312111098765432

002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)

Figure 4 Return comparing of two stocks with the CSI in the Cperiod


minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI

Figure 5 Return comparing of two stocks with the CSI in the Dperiod


Table 7 The recommended five stocks in the D period


Sunway Communicate (300136)

2013-06-17 18230 16230 2001 734 465702013-06-18 9659 8967 692 487 157902013-06-19 5176 4194 982 1137 224202013-06-20 16420 13100 3315 1330 765702013-06-21 13610 13720 minus118 minus056 minus1557Mean 12619 11242 1374 726 31959

Huafon Spandex (002064)

2013-06-17 109700 69320 40390 2144 2671102013-06-18 82970 75960 7012 440 486902013-06-19 35600 30580 5022 592 343802013-06-20 28210 28760 minus549 minus082 minus38742013-06-21 16840 15150 1688 333 11430Mean 54664 43954 10713 685 71547

Tianshan Bio-Engineer (300313)


Newcapec (300248)

2013-06-17 4379 2285 2094 1256 264702013-06-18 4035 2307 1728 1114 219902013-06-19 3748 3360 388 258 51342013-06-20 5367 6210 minus843 minus435 minus100702013-06-21 3932 4601 minus668 minus483 minus7932Mean 4292 3753 540 342 7118

Unisplendour Guhan (000590)

2013-06-17 13440 14950 minus1515 minus145 minus132802013-06-18 31530 15330 16200 1328 1677502013-06-19 6276 6254 022 004 5042013-06-20 8421 6316 2105 403 212602013-06-21 2958 6509 minus3551 minus631 minus35350Mean 12525 9872 2652 192 28177

Figure 5 shows only 19 tradingweeks Observing the return ofthe two stocks during the 20 weeks and comparing with theCSI over the same period the two stocks get the better returnsand it is far better than the CSIThe stock of Victory Precision(002426) gets the 10796 best returns at the 17th tradingweek in Figure 4 and the stock of Newcapec (300248) getsthe 4921 best returns at the 10th trading week in Figure 5Certainly we cannot get the best return in practice but theexperienced investors will get far better returns than the CSIby purchasing the recommended stocks

Figure 6 shows that different stock recommendationmodels will bring different stock returns We compute themean return rate of 18 trading weeks for 20 stocks in two dif-ferent stocks sets which were recommended by the proposedmethod and the traditionalmethod in the above four differentperiods and Figure 6 shows the result Assuming that thetarget user bought the frontal stocks in the recommend stocksset we find that the target user with appropriate operatingwill get far better stock return rate as the top line indicated inFigure 6

Number of trading week

The proposed methodThe traditional method

1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()

Figure 6 Mean return rate comparing of two different methods

In Section 41 we selected data at four different periodsof the real stock market in China Each of the periods isvery small and has only five trading days What is the reasonfor validating the recommended stocks in experimentation


Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974


on larger period with one to eighteen weeks in Figures 4and 5 As been well known it is easy for buying stockswith big order but it is very difficult for selling stocks withbig order The investors which can buy or sell stocks aregenerally institutional investors The institutional investorsmainly refer to a number of financial institutions includingbanks insurance companies investment trust companiescredit unions national bodies to establish a pension fund andother organizations

It is very difficult or impossible that the institutionalinvestors can sell out the stocks within a few days after buyinglots of stocks with big order There may be two cases Thefirst case is that the institutional investors want to sell outall stocks in the next period of day In general selling out allthe stocks required time is two to four weeks depending onthe amount of their holding stocks which is generally morethan the amount of buying stocks in those days we selectdata The other case is that the institutional investors wantto buy and not to sell some stocks in the next period of dayFor example this day when we select data is the startingperiod of buying stocks for the institutional investors Aftersome days the amount of their buying stocks reached theirdesired amount and the institutional investors are no longerbuying stocks They want to sell out all the stocks when thestocks reach a certain range of price In general the timeis twelve to sixteen weeks from buying stocks to selling outstocks for the institutional investors In fact nobody knowswhen they buy and when to sell out In order to validate theproposal we make the experimentation at the large periodfrom one to eighteen weeks after our recommend date Aswe see in Figures 4 and 5 the stock of Victory Precision(002426) and the stock of Kingsun Optoelectronic (002426)get respectively the best returns at the 17th trading weekand at the 14th trading week in Figure 4 and the stock ofNewcapec (300248) gets the better returns at the 4th tradingweek and it gets the best returns at the 10th trading week inFigure 5 Certainly we cannot get the best return in practicebecause nobody knows the highest price of the stocks incertain period and when they ought to sell out the stocks

5 Conclusions and Future Works

In this paper we proposed an efficient stock recommendationmodel based on big order net inflow The proposed stockrecommendation model based on big order net inflow cannot only filter some low investment value stocks and improvethe prediction accuracy in order to recommend some moreinvestment value stocks to the target users and get more stockreturns but also improve the speed of compute by using someadvanced algorithms such as Binary Search Algorithm andsatisfy the investment desire of the investors in real life Fromthe experimental results we found that the proposed stockrecommendation model has a better performance than themodel based on the money flow and then it can filter thelow or negative investment returns stocks and improve theinvestment gains for target investors

In the future the next work is to study how to improvethe accuracy of recommendation model We will reduce the

impact of the model on the influence of corporate events andrumors of trader and apply the model to more stock marketsor futuremarkets in the different countryWe think that thereare three aspects worth studying Firstly we can improvemodel performance with machine learning algorithms timeseries analysis or latest technology of fuzzy cluster and so onFor example wemay improve the accuracy thought training alarge number of historical stocks data sets fromdifferentmar-kets and countries using the machine learning algorithmsSecondly we may improve the accuracy of the clustering setfor stocks or users through selecting appropriate attributes ofthem or adding weights for every attribute Finally we maystudy how to find the big order transactions in real tradingprocess because some institutional investors may subdividethe big order into many random small orders by quantitativetrading system software it is very difficult for us

Conflict of Interests

All the authors of this paper declare that they have no conflictof interests in connection with the work submitted

Acknowledgments

This work is supported by the National Natural ScienceFoundation of China (no 61370073) the National HighTechnology Research and Development Program of China(no 2007AA01Z423) Sichuan Province Science andTechnol-ogy Support Program (no 2013GZX0165) the ConstructingProgram of the Key Discipline in Huaihua University theScientific Research Fund of Hunan Provincial Education(nos 12C0840 and 14C0886) the Scientific Research Fund ofHuaihua University (nos HHUY2012-15 and HHUY2011-17)the Science and Technology Plan Projects of Huaihua Cityand the Key Laboratory of Intelligent Control Technology forWuling-Mountain Ecological Agriculture in Hunan Province(no ZNKZ2014-9)

References

[1] J Duan H Liu and J Zeng ldquoPosterior probability model forstock return prediction based on analystrsquos recommendationbehaviorrdquo Knowledge-Based Systems vol 50 pp 151ndash158 2013

[2] E J de Fortuny T de Smedt D Martens and W DaelemansldquoEvaluating and understanding text-based stock price predic-tionmodelsrdquo Information ProcessingampManagement vol 50 no2 pp 426ndash441 2014

[3] L Huo B Jiang T Ning and B Yin ldquoA BP neural networkpredictor model for stock pricerdquo in Intelligent ComputingMethodologies D S Huang K H Jo and LWang Eds LectureNotes in Computer Science pp 362ndash368 Springer BerlinGermany 2014

[4] A A Adebiyi A O Adewumi and C K Ayo ldquoComparison ofARIMA and artificial neural networks models for stock pricepredictionrdquo Journal of Applied Mathematics vol 2014 ArticleID 614342 7 pages 2014

[5] K Miwa and K Ueda ldquoSlow price reactions to analystsrsquorecommendation revisionsrdquoQuantitative Finance vol 14 no 6pp 993ndash1004 2014


[6] T Geva and J Zahavi ldquoEmpirical evaluation of an automatedintraday stock recommendation system incorporating bothmarket data and textual newsrdquo Decision Support Systems vol57 no 1 pp 212ndash223 2014

[7] C S Han and C Y Wan ldquoFuture stock price predicting systemfor use in enterprise has future prediction compute serversetting up weighted value and producing virtual total amountof market price information based on basis predictive modelrdquoKR2014120416-A to Cs Co Ltd Korea Advanced Institute ofScience amp Technology 2014

[8] N C Brown K D Wei and R Wermers ldquoAnalyst recommen-dations mutual fund herding and overreaction in stock pricesrdquoManagement Science vol 60 no 1 pp 1ndash20 2014

[9] M-Y Chen ldquoA high-order fuzzy time series forecasting modelfor internet stock tradingrdquo Future Generation Computer Sys-tems vol 37 pp 461ndash467 2014

[10] B Q Sun H F Guo H R Karimi Y Ge and S XiongldquoPrediction of stock index futures prices based on fuzzy sets andmultivariate fuzzy time seriesrdquo Neurocomputing vol 151 no 3pp 1528ndash1536 2015

[11] J Patel S Shah P Thakkar and K Kotecha ldquoPredicting stockand stock price index movement using trend deterministic datapreparation and machine learning techniquesrdquo Expert Systemswith Applications vol 42 no 1 pp 259ndash268 2015

[12] D Sonsino andT Shavit ldquoReturn prediction and stock selectionfrom unidentified historical datardquoQuantitative Finance vol 14no 4 pp 641ndash655 2014

[13] M-Y Chen and B-T Chen ldquoA hybrid fuzzy time series modelbased on granular computing for stock price forecastingrdquoInformation Sciences vol 294 pp 227ndash241 2015

[14] Z Xin Z Ma and M Gu ldquoFuzzy clustering collaborativerecommendation algorithms served for directional informationrecommendationrdquoComputer Science vol 34 no 9 pp 128ndash1302007

[15] L Chapple and J E Humphrey ldquoDoes board gender diversityhave a financial impact Evidence using stock portfolio perfor-mancerdquo Journal of Business Ethics vol 122 no 4 pp 709ndash7232014

[16] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965

[17] Q Xu ldquoContinuous time MG1 queue with multiple vacationsand server close-down timerdquo Journal of Computational Informa-tion Systems vol 3 no 2 pp 753ndash757 2007

[18] A Frazzini and O A Lamont ldquoDumb money mutual fundflows and the cross-section of stock returnsrdquo Journal of Finan-cial Economics vol 88 no 2 pp 299ndash322 2008

[19] I Gath and E Bar-On ldquoComputerized method for scor-ing of polygraphic sleep recordingsrdquo Computer Programs inBiomedicine vol 11 no 3 pp 217ndash223 1980

[20] M Hong-Wei Z Guang-Wei and L Peng ldquoSurvey of collabo-rative filtering algorithmsrdquoMini-Micro Systems vol 7 pp 1282ndash1288 2009

[21] B Sarwar G Karypis J Konstan and J Riedl ldquoItem-based col-laborative filtering recommendation algorithmsrdquo inProceedingsof the 10th International World Wide Web Conference pp 285ndash295 May 2001

[22] D Goldberg D Nichols B M Oki and D Terry ldquoUsingcollaborative filtering to weave an information tapestryrdquo Com-munications of the ACM vol 35 no 12 pp 61ndash70 1992

[23] H J Ahn ldquoA new similarity measure for collaborative filteringto alleviate the new user cold-starting problemrdquo InformationSciences vol 178 no 1 pp 37ndash51 2008

[24] G Choudhury ldquoSome aspects of MG1 queue with two differ-ent vacation times under multiple vacation policyrdquo StochasticAnalysis and Applications vol 20 no 5 pp 901ndash909 2002

[25] B T Doshi ldquoConditional and unconditional distributions forMG1 type queues with server vacationsrdquo Questa vol 7 pp229ndash252 1990

[26] H W Lee ldquoMG1 queue with exceptional first vacationrdquoComputers amp Operations Research vol 15 no 5 pp 441ndash4451988

[27] K K Leung ldquoOn the additional delay in an MG1 queuewith generalized vacations and exhaustive servicerdquo OperationsResearch vol 40 supplement 2 pp S272ndashS283 1992

[28] A Hatamlou ldquoIn search of optimal centroids on data clusteringusing a binary search algorithmrdquo Pattern Recognition Lettersvol 33 no 13 pp 1756ndash1760 2012

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of


Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of


Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of


Mathematical PhysicsAdvances in

Complex AnalysisJournal of


OptimizationJournal of


CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of


Operations ResearchAdvances in

Journal of


Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences


The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Algebra

Discrete Dynamics in Nature and Society



Decision SciencesAdvances in

Discrete MathematicsJournal of


Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of


prerecommended stock set Then we use a method based onMG1 [17] to compute the net inflow amount of big orderfor every stock in the prerecommended stock set From theprerecommended stock set descending ordered by the valueof big order net inflow we choose some stocks with highestvalue in front of the set as the last recommendation stock setfor the target user

In general we believe that the stock rise with a largetrading volume is closely related to the purchase stock ofbig order In order to analyze the big order net flow ofstock we need to observe the stock trading volume andturnover The money net flow and the money flow [18] aredifferent in concept The big order refers to the amount ofeach transaction over one million yuan or the volume of eachtransaction over fifty thousand in a single transaction So thebig order net inflow refers to the amount of money of bigorder buy or sell of the same stocks within a day Most ofthe time the money flow is bigger than zero and the moneynet flow is less than zero In individual cases the money netflow is bigger than zero and the big order net flow is lessthan zero Under this situation we proposed an efficient stockrecommendation model based on big order net inflow Thenew recommendation model can measure the capital andthe pulsation of the stock markets and consider investorspreferences and behavior characteristics it can improve theexisting deficiencies of some current stock recommendationIn addition the new recommendationmodel can analyze andfilter the stock with less returns in the future and improvethe investment gains of investors The experiment resultsshow that the recommended stocks have better gains duringthe several days after the recommended stock day and theproposedmodel can provide reliable investment guidance forthe target investors and let them getmore investment returns

The rest of this paper is organized as follows Section 2briefly reviews the definitions and theorems of fuzzy cluster-ing and the framework of the collaborative filtering algorithmbased on user fuzzy clustering Section 3 demonstrates themethod for computing big order net inflow and the frame-work of the proposed model Section 4 presents the simu-lation experiment and empirical analyses of the proposedmodel finally some conclusions are given and some futureworks are pointed out in Section 5

2 Theoretical and Modeling Framework

The concept of fuzzy set [16] in fuzzy cluster is put forward byZadeh in 1965 Gath and Bar-On earlier applied that theoryto compute the scoring of poly graphic sleep recordings intheir study [19] In this section we will briefly review thedefinitions and theoretic of fuzzy cluster

21 Fuzzy Clustering Theory


a matrix with 119898 rowsand 119899 columns if 0 le 119903

119894119895le 1 then 119877 is called fuzzy matrix

119877 When 119903119894119895is only 0 or 1 said 119877 is a Boolean matrix When

elements 119903119894119894of the diagonal are all 1 in fuzzy matrix 119877 =

(119903119894119895)119899times119899

said 119877 is reflexive fuzzy matrix

Definition 2 If119860 is 119899-order square defined1198602= 119860 ∘1198601198603 =

1198602∘ 119860 and 119860

119896= 119860119896minus1

∘ 119860


and called 119877119879

=

(119903119879

119894119895)119898times119899

is the transposed matrix of 119877 where 119903119879119894119895= 119903119894119895


for any 120582 isin [0 1] andsaid 119860

120582= (119903(120582)

119894119895)119898times119899

is the 120582-cut matrix of the fuzzy matrixR When 119903

119894119895ge 120582 119903(120582)

119894119895= 1 and 119903

119894119895lt 120582 119903(120582)

119894119895= 0 said 119860

120582is a

Boolean matrix and said 120582 is a confidence level parameter orcut level parameter

Definition 5 Suppose two limited discourse domains 119883 =

1199091 1199092 119909

119898 and 119884 = 119910

1 1199102 119910

119899 then 119883 to 119884 fuzzy

relation 119877 is an119898times119899 order fuzzy matrix and set 119877 = (119903119894119895)119898times119899

where 119903

119894119895= 119877(119909

119894 119910119895) isin [0 1] represents (119909

119894 119910119895) relevance on

fuzzy relation 119877

Theorem 6 If 119877 is fuzzy similar matrix then for any naturalnumber 119896 119877119896 is fuzzy similarity matrix

Theorem 7 If 119877 is an 119899 order fuzzy similar matrix then thereis minimum natural number 119896 (119896 le 119899) for all natural number119898 (119898 ge 119896) constant 119877119898 = 119877

119896 119877119896 namely fuzzy equivalentmatrix (1198772119896 = 119877

119896) At this point said119877119896 is the transitive closure

of 119877 denoted by 119905(119877) = 119877119896

Method 8 If 119877 is fuzzy similar matrix then the followingmethod will solve the transitive closure 119905(119877) 119877 rarr 119877

2rarr

1198774rarr 1198778rarr 11987716

rarr sdot sdot sdot rarr 1198772119896

and the said method is thesquare self-synthesis method (S3M)

22 Fuzzy Clustering Analysis The fuzzy clustering analysisis an analyzing clustering and classification method by estab-lishing fuzzy similar relationship of objective things basedon the objective characteristics the degree of closeness andsimilarity between objective things

In Figure 1 the fuzzy clustering processing model can bedivided into four stages namely the data preprocessing thedata standardization the constructing fuzzy similar matrix(FSM) and the clustering and analysis

221 Data Preprocessing In China it is well known thatindividual investors buy or sell stocks only through a secu-rities company not directly from the stock exchange Inorder to trade stocks individual investors firstly have toregister as a member in a securities company Accordingto the National Security Act investors must have a testin the risk tolerance when investors are registering Thesecurities company gives them a risk test paper with fifteenquestions or more Each question has four or five options forinvestors to choose In order to store and process we usean integer value instead of the chosen answer of investorsin the user risk database In this paper we choose twelveanswers among all answers to study There are at least 1million records about risk information of investors in theuser risk database of any securities company because there



119909119894119895














Clustering




Constructing FSM




Standardization




Preprocessing






1 1199092 119909

119899 where 119909

119894



119894= 1199091198941 1199091198942 1199091198943 119909

119894119895 119909




119884 = 119910119894119895 1 le 119894 le 119899 1 le 119895 le 12

119910119894119895=

119909119894119895minusmin

1le119894le119899119909119894119895

max1le119894le119899

119909119894119895 minusmin

1le119894le119899119909119894119895

1 le 119895 le 12

119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


1199101198991

1199101198992


]]]]]]

]

(1)


119903119894119895= 1 minus 119888 times 119889 (119909

119894 119909119895) (2)



119894and 119909

119895as

follows

119889 (119909119894 119909119895) = radic

12

sum

119896=1

(119909119894119896minus 119909119895119896)2

119888 =1

1 + 1198881015840

(3)





119877 =

[[[[[[

[

11990311

11990312


11990321

11990322


1199031198991

1199031198992

sdot sdot sdot 119903119899119899

]]]]]]

]

(4)



119905 (119877) 119877 997888rarr 1198772997888rarr 119877

4997888rarr 119877

8997888rarr 119877


997888rarr 1198772119896

(5)

where 119896 le [log1198992]










1 1198802 1198803 119880



119894) ordered by

descending



119891 (119906 119895) =

radicsum119870

119896=1sim (119894 119896)

2lowast (119878119896minus 119878119896)2

radicsum119870

119896=1sim (119894 119896)

2

+ 119878119906 (6)






sim (119894 119896) =

sum119898

V=1 (119877V119894 minus 119877119894) (119877V119896 minus 119877

119896)

radicsum119898


V=1 (119877V119896 minus 119877119896)2

(7)


119896is










0 = 119879119894

lt 119878119894

lt 119879119894+1

lt 119878119894+1


lt 119878119899 We





= 119908(119878119894) and 119882

2119894= 119908(119879

119894+1)





119898

119894=1(119878119894minus 119879119894) ge 119896 and 119864(119896) =

sum119871(119896)minus1

119894=1(119879119894+1

minus 119878119894)

If 119896 lt sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896) = 119908(119896 + 119864(119896))

If 119896 = sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896

+) = 119908(119896 + 119864(119896) + 119879

119871(119896)+1lt

119878119871(119896)

) and 119897(119896minus) = 119908(119896 + 119864(119896))

Clearly if 119896 = sum119871(119896)

119894=1(119878119894minus119879119894) then 119897(119896

+) minus 119897(119896

minus) = 119882

2119871(119896)minus

1198821119871(119896)


3 Proposed Method

















119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)


119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)




119899be the number




119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)


119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)


119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)


119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot


))))))

)

(14)


119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)



User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra











119909119894119895














Clustering




Constructing FSM




Standardization




Preprocessing






1 1199092 119909

119899 where 119909

119894



119894= 1199091198941 1199091198942 1199091198943 119909

119894119895 119909




119884 = 119910119894119895 1 le 119894 le 119899 1 le 119895 le 12

119910119894119895=

119909119894119895minusmin

1le119894le119899119909119894119895

max1le119894le119899

119909119894119895 minusmin

1le119894le119899119909119894119895

1 le 119895 le 12

119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


1199101198991

1199101198992


]]]]]]

]

(1)


119903119894119895= 1 minus 119888 times 119889 (119909

119894 119909119895) (2)



119894and 119909

119895as

follows

119889 (119909119894 119909119895) = radic

12

sum

119896=1

(119909119894119896minus 119909119895119896)2

119888 =1

1 + 1198881015840

(3)





119877 =

[[[[[[

[

11990311

11990312


11990321

11990322


1199031198991

1199031198992

sdot sdot sdot 119903119899119899

]]]]]]

]

(4)



119905 (119877) 119877 997888rarr 1198772997888rarr 119877

4997888rarr 119877

8997888rarr 119877


997888rarr 1198772119896

(5)

where 119896 le [log1198992]










1 1198802 1198803 119880



119894) ordered by

descending



119891 (119906 119895) =

radicsum119870

119896=1sim (119894 119896)

2lowast (119878119896minus 119878119896)2

radicsum119870

119896=1sim (119894 119896)

2

+ 119878119906 (6)






sim (119894 119896) =

sum119898

V=1 (119877V119894 minus 119877119894) (119877V119896 minus 119877

119896)

radicsum119898


V=1 (119877V119896 minus 119877119896)2

(7)


119896is










0 = 119879119894

lt 119878119894

lt 119879119894+1

lt 119878119894+1


lt 119878119899 We





= 119908(119878119894) and 119882

2119894= 119908(119879

119894+1)





119898

119894=1(119878119894minus 119879119894) ge 119896 and 119864(119896) =

sum119871(119896)minus1

119894=1(119879119894+1

minus 119878119894)

If 119896 lt sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896) = 119908(119896 + 119864(119896))

If 119896 = sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896

+) = 119908(119896 + 119864(119896) + 119879

119871(119896)+1lt

119878119871(119896)

) and 119897(119896minus) = 119908(119896 + 119864(119896))

Clearly if 119896 = sum119871(119896)

119894=1(119878119894minus119879119894) then 119897(119896

+) minus 119897(119896

minus) = 119882

2119871(119896)minus

1198821119871(119896)


3 Proposed Method

















119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)


119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)




119899be the number




119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)


119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)


119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)


119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot


))))))

)

(14)


119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)



User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra













119877 =

[[[[[[

[

11990311

11990312


11990321

11990322


1199031198991

1199031198992

sdot sdot sdot 119903119899119899

]]]]]]

]

(4)



119905 (119877) 119877 997888rarr 1198772997888rarr 119877

4997888rarr 119877

8997888rarr 119877


997888rarr 1198772119896

(5)

where 119896 le [log1198992]










1 1198802 1198803 119880



119894) ordered by

descending



119891 (119906 119895) =

radicsum119870

119896=1sim (119894 119896)

2lowast (119878119896minus 119878119896)2

radicsum119870

119896=1sim (119894 119896)

2

+ 119878119906 (6)






sim (119894 119896) =

sum119898

V=1 (119877V119894 minus 119877119894) (119877V119896 minus 119877

119896)

radicsum119898


V=1 (119877V119896 minus 119877119896)2

(7)


119896is










0 = 119879119894

lt 119878119894

lt 119879119894+1

lt 119878119894+1


lt 119878119899 We





= 119908(119878119894) and 119882

2119894= 119908(119879

119894+1)





119898

119894=1(119878119894minus 119879119894) ge 119896 and 119864(119896) =

sum119871(119896)minus1

119894=1(119879119894+1

minus 119878119894)

If 119896 lt sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896) = 119908(119896 + 119864(119896))

If 119896 = sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896

+) = 119908(119896 + 119864(119896) + 119879

119871(119896)+1lt

119878119871(119896)

) and 119897(119896minus) = 119908(119896 + 119864(119896))

Clearly if 119896 = sum119871(119896)

119894=1(119878119894minus119879119894) then 119897(119896

+) minus 119897(119896

minus) = 119882

2119871(119896)minus

1198821119871(119896)


3 Proposed Method

















119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)


119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)




119899be the number




119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)


119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)


119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)


119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot


))))))

)

(14)


119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)



User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra














sim (119894 119896) =

sum119898

V=1 (119877V119894 minus 119877119894) (119877V119896 minus 119877

119896)

radicsum119898


V=1 (119877V119896 minus 119877119896)2

(7)


119896is










0 = 119879119894

lt 119878119894

lt 119879119894+1

lt 119878119894+1


lt 119878119899 We





= 119908(119878119894) and 119882

2119894= 119908(119879

119894+1)





119898

119894=1(119878119894minus 119879119894) ge 119896 and 119864(119896) =

sum119871(119896)minus1

119894=1(119879119894+1

minus 119878119894)

If 119896 lt sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896) = 119908(119896 + 119864(119896))

If 119896 = sum119871(119896)

119894=1(119878119894minus 119879119894) then 119897(119896

+) = 119908(119896 + 119864(119896) + 119879

119871(119896)+1lt

119878119871(119896)

) and 119897(119896minus) = 119908(119896 + 119864(119896))

Clearly if 119896 = sum119871(119896)

119894=1(119878119894minus119879119894) then 119897(119896

+) minus 119897(119896

minus) = 119882

2119871(119896)minus

1198821119871(119896)


3 Proposed Method

















119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)


119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)




119899be the number




119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)


119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)


119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)


119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot


))))))

)

(14)


119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)



User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra






















119875 (119905119894ge 119905) = 119861 (119905) 119894 = 1 2 3 (8)


119883 =1

119906= int

infin

0

119905119889119861 (119905)

1198832 = int

infin

0

1199052119889119861 (119905)

(9)




119899be the number




119899 then

119883119899+1

= 119883119899+ 119860 (119889

119899+1) minus 119860 (119889

119899) minus 1 (10)

or

119889119899+1

= 119889119899+ 119905119899+ 1 (11)


119883119899+1

= 119860 (119886119899+1

+ 119905119899+1

) minus 119860 (119886119899+1

) (12)


119875 119883119899+1

= 119894 | 1198831 1198832 119883

119899 = 119875 119883

119899+1= 119894 | 119883

119899 (13)


119875 =

((((((

(

1199010

1199011

1199012

1199013

sdot sdot sdot

1199010

1199011

1199012

1199013

sdot sdot sdot

0 1199010

1199011

1199012

sdot sdot sdot

0 0 1199010

1199011

sdot sdot sdot


))))))

)

(14)


119896is given by

119901119896= int

infin

0

119890minus120582119905

(120582119905)119896

119896119889119861 (119905) (119896 = 0 1 2 ) (15)



User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra










User clustering

Stock recommend









(a)


Counting Counting

Clustering

Target user

Finding

User risk DB

Usercategories


Neighbor userlist



Finding


Stock DB

Stock categories

Clustering

(b)





247068

237898

228603

219433

210138

200968

191673A

B

C

D

244480

184965













2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra


















2clustered









Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra














Mean 53683 27140 26542 1086 144982



Mean 262 080 182 569 1926



Mean 16763 14267 2494 467 45840



Mean 2475 1802 673 202 4140

Faw Car (000800)


Mean 12400 9860 2540 038 15992







2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra













2012-12-17 176520 101220 75300 1659 5306902012-12-18 101830 83250 18590 512 1280602012-12-19 120750 72390 48360 1988 3346802012-12-20 139420 85260 54160 2045 3766902012-12-21 59740 46350 13390 617 95600Mean 119652 77694 41960 1364 293144


















Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra














Mean 1729 1281 448 940 6837



Mean 2350 1915 435 472 5214



Mean 2001 1670 330 417 5848



Mean 25578 23710 1868 352 19714



Mean 3366 2378 988 347 6670


002638002426CSI

minus40

minus20

0

20

40

60

80

100

120

Retu

rn ra

te (

)



minus20

minus10

0

10

20

30

40

50

60

Retu

rn ra

te (

)

300248000590CSI











Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra


















Newcapec (300248)








1 18171615141312111098765432minus10

0

10

20

30

40

50

Mea

n re

turn

rate

()




Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra










Table8Th

erecom

mendedsto

cksm

axim

umpo

ssiblegainsd

uringthefou

rdifferentd

ayintervals

Recommend

dateperiod

Stockname(code)CS

IRe

commend

buyop

enPrice

Highestprice

with

in10

days

Rise-up

with

in10

days

Highestprice

with

in30

days

Rise-up

with

in30

days

Highestprice

with

in90

days

Rise-up

with

in90

days

Highestprice

with

in1y

ear

Rise-upwith

in1y

ear

2012-12-05

Aperio

d

Jiang

huaiAu

tomob

ile(600

418)

556

601

809

698

2554

939

6888

1037

8651

Janu

sPrecisio

n(300

083)

1014

1160

1440

1255

2377

1735

7110

2500

14655

GreatWallM

otor

(601633)

1843

2080

1286

2391

2973

3150

7092

5285

18676

HuasunGroup

(000790)

585

667

1402

710

2137

855

4615

1289

12034

FawCa

r(00

0800)

630

713

1317

872

3841

879

3952

1588

15206

Shangh

aiCom

positeInd

ex(C

SI)

197311

215250

909

2296

111637

2444

80

2391

2444

80

2391

2012-12-24

Bperio

d

DahuAq

uacultu

re(600257)

683

742

864

74

2864

74

2864

93

63704

ChanganAu

tomob

ile(000

625)

617

673

908

762

2350

956

5494

1319

11378

Xinn

ingMod

ern(300

013)

920

1143

2424

1143

2424

1143

2424

1997

11707

Robam

Appliances

(002508)

1862

1935

392

2500

3426

2687

4431

4279

1298

1GreatWallM

otor

(601633)

2245

2391

650

3039

3537

3418

5225

5285

13541

Shangh

aiCom

positeInd

ex(C

SI)

2150

65

2296

11676

233581

861

2444

80

1368

2444

80

1368

2013-02-18

Cperio

d

IrtouchSyste

ms(300282)

1555

1558

019

1661

682

1661

682

3269

11023

Transin

foTechno

logy

(002373)

1190

1222

269

1273

697

1401

1773

1950

6387

MicrogateTechno

logy

(300319)

1765

1862

550

1866

572

1866

572

3763

11320

King

sunOptoelectronic

(002638)

1080

1265

1713

1319

2213

1542

4278

1620

5000

VictoryPrecision

(002426)

678

722

649

722

649

1250

8437

2046

2017

7Shangh

aiCom

positeInd

ex(C

SI)

2444

80

2444

80

000

24

4480

000

24

4480

000

24

4480

000

2013-06-24

Dperio

d

Sunw

ayCom

mun

icate(300136)

2305

2720

1800

2720

1800

2720

1800

2766

2000

HuafonSpandex(00206

4)664

773

1642

830

2500

1150

7319

1150

7319

Tianshan

Bio-En

gineer

(300313)

1341

1444

768

1444

768

1663

2401

2310

7226

New

capec(300248)

1199

1485

2385

1790

4929

1919

6005

3132

1612

2Unisplend

ourG

uhan

(000590)

990

938

minus525

939

minus515

1350

3636

1779

7970

Shangh

aiCom

positeInd

ex(C

SI)

2068

86

2068

86

000

20

9287

116

2270

27

974

2270

27

974










Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra


















Acknowledgments


References





































Volume 2014




Journal of











Journal of


Function Spaces






Algebra








































Volume 2014




Journal of











Journal of


Function Spaces






Algebra









Research Article An Efficient Stock Recommendation Model...

Documents

Transcript of Research Article An Efficient Stock Recommendation Model...