Smart data sumit DXB 2015: 80/20 for Airlines

THE 80/20 RULE IN BIGDATA

BY JOSE BERENGUERES & DMITRY EFIMOV

Smart Data Summit Dubai | 25 - 26 May 2015

How to optimise your Loyalty program from a non-value-added-activity point of

view and other stuff

case based on: Airline new customer tier level forecasting for real-time resource allocation of a miles program http://www.journalofbigdata.com/content/1/1/3

BACKGROUND: ROBOTICS

BACKGROUND: DESIGN THINKING

BACKGROUND: BIOINSPIRED MANUFACTURING

Loyalty miles program (15,000,000 rows)

demographics, miles, flights

?

Linear Forecasting (42%)

80/20

Today ?

Today ?

Rank, Resource optimisation

Testing the limits: 2 weeks

ordered by miles max to min.

HOW IT WAS DONE…

1.Time Shift CRM events to present*

2.Feature Extraction (Cluster

Dummy Method)

3.Generalized Boosting Machine

(Tree Search)

4.Generalized Linear Model

(Regression based)

5.Blending with grid search (+5%)

*the computer understand dates but it understands better them in terms of “how many days ago”

Conclusion

Software Lock-ins

Asking the right question is the hard part of big data

Huge opportunities in 80/20: Waste optimisation