Ppt

29
LINK PREDICTION IN SOCIAL NETWORKING SUBMITTED BY: Umang Chaudhary (10103408) Sanyam Gupta (10103405) Under the Guidance Of: Dr. Buddha Singh

description

this is a final year project on link prediction in social networking where users are predicted on the mentioned algorithm in the slide

Transcript of Ppt

Page 1: Ppt

LINK PREDICTION IN SOCIAL NETWORKING

SUBMITTED BY:Umang Chaudhary (10103408)Sanyam Gupta (10103405)

Under the Guidance Of: Dr. Buddha Singh

Page 2: Ppt

WHAT IS SOCIAL NETWORKING?

Social networking is the grouping of individuals into specific groups, like small rural communities or a neighbourhood subdivision, if you will. Although social networking is possible in person, especially in the workplace, universities, and high schools, it is most popular online.

Page 3: Ppt

CHALLENGE…!!Social networks are highly dynamic objects;

they grow and change quickly over time through the addition of new edges. Its solution is LINK PREDICTION.

Page 4: Ppt

WHAT IS LINK PREDICTION?

Link prediction is an important task for analysing social networks. It is the method to predict link between the given nodes using various algorithms.

Page 5: Ppt

APPLICATIONSIdentifying the structure of a criminal

network by predicting missing links in a criminal network using incomplete data.

Automatic web hyperlink creation

 Website hyperlink prediction

Build recommendation systems (e-commerce)

Page 6: Ppt

Protien-protien interaction (bio-informatics)

Annotate PPI graph (bio-informatics)

Identify hidden group of terrorists (security)

Overcoming the data-sparsity problem in recommender systems using collaborative filtering.

Page 7: Ppt

PROBLEM STATEMENTThe network is dynamic as it

keeps on expanding because the users keep getting added exponentially so predicting link between the users is a big challenge. So we are going to implement algorithms like common neighbours, jaccard coefficient and adamic/adar which can predict link efficiently,

Page 8: Ppt

Our social networking application is about where anyone can register and become a member and stay connected to all their friends and to other users. Our application includes all the features of a social networking application like:

We have developed the social networking system which is client-server based.

We have developed the algorithms for link prediction that includes prediction on the basis of common neighbours, common features, members of the same community etc.

Page 9: Ppt

To chek user’s strength we have calculated betweeness, closeness and degree of centrality.

We have implemented our proposed algorithm using python programming language, GLADE(for GUI) and SQL (for database).

We have evaluated the performance of our proposed algorithm.

Page 10: Ppt

FEATURES OF OUR APPLICATION New user can register on the application and can update their

profile and also can view other’s profile.  Users have a unique profile visible to the friends and users of

the application, where they can upload there pictures and personal information.

Users can add other users.

Most famous person in the group can be predicted in the application.

We are able to find each user’s strength by calculating closeness, betweeness and degree of centrality.

Users can upload their status and comment on the status as well.

Page 11: Ppt

Users can join communities created by them in their field of interest.

Users can join groups, create groups.

There is link prediction on the basis of common or mutual friends so the users will get friend suggestion on the basis of mutual friends

The users get friend suggestion also from the person who are added in the same community if there are many features common in them.

We are able to find out who have uploaded similar status (the way it is done in twitter).

Page 12: Ppt

Users also get friend suggestions of those persons who have many features in common.

Users can chat with each others.

The application contains basic features of any social networking website such as liking the pages, adding friends, uploading status, follow people etc.

Page 13: Ppt

METHODS FOR LINK PREDICTION

There can be many methods that can be used for predicting the link. Some of them are:

1. Jaccard coefficient2. Adamic/adar3. Common neighbors4. Graph distance5. Katz6. Hitting time7. Friends measure8. Preferential attachment score9. Bayesian algorithm, etc.

Page 14: Ppt

The link prediction problem

Given a snapshot of a social network at time t, we seek to accurately predict the edges that will be added to the network during the interval (t, t’)

Page 15: Ppt

Methods we are using for the prediction

COMMON NEIGHBOURS

JACCARD COEFFICIENT

ADAMIC/ADAR

Page 16: Ppt

COMMON NEIGHBOURSA and C have 2 common neighbors,

more likely to collaborate

A

B

C

D

E

Page 17: Ppt

JACCARD COEFFICIENTsame as common neighbors, adjusted

for degree

Page 18: Ppt

ADAMIC/ADARweighting rarer neighbors more

heavily (gives more weightage to neighbours that are not shared with many others)

A

B

C

DE

Page 19: Ppt

Methods we are using to find strengths of users1. Degree of Centrality: Centrality is

regarded as one of the most important and commonly used conceptual tools for exploring actor roles in social networks. A node’s degree centrality, in an un-directed graph, is defined as the number of nodes that are connected to that node.

The definition dictates that “central actorsmust be the most active in the sense that

theyhave the most ties to other actors in thenetwork or graph”

Page 20: Ppt

2. Closeness: Closeness centrality indicates the influence of a node on the entire network, and thus discipline centrality in this research can tell how “close” each discipline is to the other disciplines and the influence that a discipline puts on the entire network.

3. Betweenness: According to the definition of betweenness, betweenness centrality reflects the bridge role of a discipline in a knowledge communication network. The larger the discipline betweenness, the more control that the discipline has over the interaction between other disconnected disciplines.

Page 21: Ppt

IMPLEMENTATIONSANPSHOTS

Page 22: Ppt
Page 23: Ppt
Page 24: Ppt
Page 25: Ppt

TEST PLANWe have implemented many test cases

on the modules that have been developed so far which are as follows:

When we check the prediction on the basis of common features, we see that the right person gets predicted.

On changing the features of any users, new links should pe predicted which is as per the adamic and adar algorithm.

Page 26: Ppt

Testing the user login by entering the user name and password. If the user name and password is correct then the user will login otherwise login failed. Below are the snapshots indicating the login successful and failure.

If all the fields have been filled for registering the user, then the user will be registered with a different id which is the primary key.

 On adding friends we test that the friend list

should get updated successfully.

Page 27: Ppt

REFERENCES  ULRIK BRANDES†.2001. A Faster Algorithm For Betweenness Centrality*. Published in Journal of

Mathematical Sociology

Kazuya Okamoto1, Wei Chen, and Xiang-Yang Li.2008. Ranking of Closeness Centrality for Large-

Scale Social Networks. FAW '08 Proceedings of the 2nd annual international workshop on Frontiers

in Algorithmics ( Pages 186-195)

Luca Maria Aiello, Alain Barrat, Rossano Schifanella, Ciro Cattuto, Benjamin Markines, Filippo

Menczer. 2011. Friendship prediction and homophily in social media. ACM journal.

Purnamrita Sarkar, Deepayan Chakrabartiy, Michael I. Jordanz. 2012. Nonparametric Link

Prediction in Dynamic Networks. ICML, UK,2012, PAGE NO. 1-8

Zhengdong Lu, Berkant Savas, Wei Tang, Inderjit Dhillon. 2010. Supervised Link Prediction Using

Multiple Sources. 29 th International Conference on Machine Learning, Edinburgh, Scotland, UK.

Suphakit Niwattanakul, Jatsada Singthongchai, Ekkachai Naenudorn and Supachanun

Wanapu.2013. Using of Jaccard Coefficient with keyword similarity. IMECS

AIROLDI, E., BLEI, D., FIENBERG, S., XING, E., AND JAAKKOLA, T. 2006. Mixed membership

stochastic block models for relational data with application to protein-protein interactions. In

Proceedings of the international biometrics society annual meeting.

Page 28: Ppt

AVIN, C., LOTKER, Z., AND PIGNOLET, Y. 2011. On the elite of social networks. Arxiv preprint

arXiv:1111.3374.

BARABASI, A.-L. AND ALBERT, R. 1999. Emergence of scaling in random networks. Science 286,

509–512.

BLONDEL, V., GUILLAUME, J., LAMBIOTTE, R., AND LEFEBVRE, E. 2008. Fast unfolding of

communities in large networks. Journal of Statistical Mechanics: Theory and Experiment 2008,

P10008.

CHA, M., HADDADI, H., BENEVENUTO, F., AND GUMMADI, K. P. 2010. Measuring user influence in

twitter: The million follower fallacy. In In Proceedings of the 4th International AAAI Conference

on Weblogs and Social Media (ICWSM). Washington DC, USA.

CHAWLA, N., JAPKOWICZ, N., AND KOTCZ, A. 2004. Editorial: special issue on learning from

imbalanced data sets. ACM SIGKDD Explorations Newsletter 6, 1, 1–6.

CHO, E., MYERS, S., AND LESKOVEC, J. 2011. Friendship and mobility: user movement in

location-based social networks. In Proceedings of the 17th ACM SIGKDD international

conference on Knowledge discovery and data mining. ACM, 1082–1090.

A.-L. Barab´asi and R. Albert, “Emergence of Scaling in Random Networks,” Science, vol. 286,

no. 5439, pp. 509-512, Oct. 1999.

D. Wang, Dashun. Pedreschi and A. Barab´asi, “Human Mobility, Social Ties, and Link

Prediction,” in KDD ’11. ACM, 2011, pp. 1100–1108.

Page 29: Ppt

THANK YOU