Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

24
Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st , 2014

description

Clustering A cluster is a small group or bunch of something. We chose k-means clustering, which aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean.

Transcript of Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Page 1: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Sabre Senior Design ProjectHUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLEMay 1st, 2014

Page 2: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

▪ Sabre Airline Solutions would like to provide traveler segmentation services for their customer reservation system to support various marketing programs.

The Problem

Page 3: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Clustering

• A cluster is a small group or bunch of something.

• We chose k-means clustering, which aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean.

Page 4: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Use Cases▪If an airline creates a new ticket fare product, but our segmentation shows a wide variety in client demographics (30% leisure, 40% business, and 30% other) then perhaps the product is not well-defined or well-targeted.

▪A good segmentation can provide insights into the design of good fare products upfront.

▪Conduct break down analysis of various fare products by post-booking and pre-booking segmentation.

Page 5: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Information Given

▪Sabre provided ticketing data in an excel spreadsheet, with rows representing each ticket:- 14 days of ticketing- 2,535,955 tickets- Booking date, departure date, return date (and day-of-week)- Fare or fare ratio (ratio to lowest fare)- Travel time or ratio (ratio to non-stop time) OR number of stops (outbound and inbound), Market, Travel agency type (big global, small, online agency)

▪Sample R code referencing the data metrics above

Page 6: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Our Goal▪ Create a set of rules that creates a unique category for any

given ticket, along with a report showing validation efforts. Our report will also describe the techniques used, pros and cons, and rationale for the final recommended set of rules.

Page 7: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Methodology»Used k-means clustering in R

»Found optimal number of clusters

»Interpreted the output

»Used subjective analysis to determine any conclusions

Page 8: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Number of Clusters vs R2

R2

Number of Clusters

Page 9: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

▪ Set of Rules to classify clusters: 1. Determine if the value displayed in the table is above or below

the mean for the following variables: Advanced Purchase, Length of Stay, Fare, Inbound Travel Time and Outbound Travel Time

2. Determine the most popular of the following variables:Departure Day of Week, Return Day of Week, Agency Type, and Number of Passengers

3. Check percentage of tickets in the cluster that stays on Saturdays (we need to determine if its high or low)

Post-booking Classification Rules

Page 10: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C1- Last Minute Single Business Traveler

Page 11: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C2- Extended Stay Leisure Traveler

Page 12: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C3- Cheap Business Traveler

Page 13: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C4- Planned Vacation Traveler

Page 14: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C5- Recurrent Business Traveler

Page 15: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

C6- Quick-Trip Business Traveler

Page 16: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

▪ Same rules as for Post-booking but with the removal of the following variables:oFareoOutbound Travel TimeoInbound Travel Time

Pre-booking Classification Rules

Page 17: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 1- Leisure Weekend

Cluster 1 – leisure weekend

• Increased Advanced Purchase and LOS

• Departs Thursday/Friday

• Return Sunday/Monday

• Gonline

• 96% Stay Saturday

• 70% Single

Page 18: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 2- Leisure Traveler

Cluster 2 – leisure travelers

• Increased Advanced Purchase and LOS

• Departs Any Day

• Return Any Day

• Gonline

• 85% Stay Saturday

• 80% Single

Page 19: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 3- Holiday/Vacation Travelers

Cluster 3 – quick-trip business travelers

• Short L.O.S.

• Departs Monday/Tuesday

• Gcorp

• 93%

Cluster 3 – holiday/vacation travelers

• Increased Advanced Purchase and LOS

• Departs Any Day

• Return Sunday

• Gonline & Gcorp

• 50% Stay Saturday

• 67% Single, 22% Couple, 11% Family

Page 20: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 4- Week-long Business Traveler

Cluster 4 – quick-trip business travelers

• Short L.O.S.

• Departs Monday/Tuesday

• Gcorp

• 93%

Cluster 4 – week-long business traveler

• Avg. Advanced Purchase and increased LOS

• Departs Sunday/Monday

• Return Thursday/Friday

• Gcorp

• 0.03% Stay Saturday

• 95% Single

Page 21: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 5- Quick-Trip Business Travelers

Cluster 5 – quick-trip business travelers

• Short L.O.S.

• Departs Monday/Tuesday

• Gcorp

• 93%

Cluster 5 – quick-trip business travelers

• Avg. Advanced Purchase and decreased LOS

• Departs Monday/Tuesday/Wednesday

• Return Thursday/Friday

• FSC and Unclass

• No Stay Saturday

• 93% Single

Page 22: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Cluster 6- Gcorp Business Travelers

Cluster 6 – quick-trip business travelers

• Short L.O.S.

• Departs Monday/Tuesday

• Gcorp

• 93%

Cluster 6 – gcorp business travelers

• Avg. Advanced Purchase and decreased LOS

• Departs Monday/Tuesday/Wednesday

• Return Thursday/Friday

• 100% Gcorp

• No Stay Saturday

• 99% Single

Page 23: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

Proposals

• Use post-booking clusters to offer discounts or premiums in post-booking services such as hotels and other reservations.o i.e. Cluster 5: Recurrent Business Travelers could receive a

discount in airport lounges or partnered hotels.

• Use pre-booking clusters to choose a more appropriate fare targeted to each cluster.o i.e. Cluster 3: Vacation travelers would prefer to pay a

cheaper fare because they book in advance.

Page 24: Sabre Senior Design Project HUNTER ROSS, RAMON TRESPALACIOS, MARY LIZ TUTTLE May 1 st, 2014.

World Changers Shaped Here