American Express Slides, MLconf 2013

  • View
    1.237

  • Download
    1

  • Category

    Business

Preview:

DESCRIPTION

 

Citation preview

MLConf, San Francisco, CA November 15, 20131

Recommendations @ American Express

Abhijit Bose, Henry H Yuan and Huiming Qu

Data Science and EngineeringAmerican Express Company

MLConf, San Francisco, CA November 15, 2013

MLConf, San Francisco, CA November 15, 2013 2MLConf, San Francisco, CA November 15, 2013

American Express Today

MLConf, San Francisco, CA November 15, 2013 3

Our closed loop gives us direct relationships with millions of buyers and sellers

and a wealth of informationabout buyers and sellers

MLConf, San Francisco, CA November 15, 2013 4

Trust and security have been the hallmarks of the American Express brand for more than 160 years.

Turning good data into more tailored and targeted commerce does not change our privacy policies and principles.

We know customers need transparency and clear explanations.

We use data to better serve our customers. We do not sell personally identifiable information in any context.

Our products must adhere to the highest standards

MLConf, San Francisco, CA November 15, 2013 6

Recommendation opportunities exist in

many different channels

MLConf, San Francisco, CA November 15, 2013 7

My Offers Mobile App

MLConf, San Francisco, CA November 15, 2013 8

https://sync.americanexpress.com/

MLConf, San Francisco, CA November 15, 2013 9

Website Personalization

MLConf, San Francisco, CA November 15, 2013 10

Merchant Insight Portal

MLConf, San Francisco, CA November 15, 2013 11

Merchant Name

Merchant Street Address

Total Amount

Amex card used

Merchant Zip Code

Transaction Timestamp

Transaction ID (useful for history, e.g. returns, tips, etc)

What a Typical Transaction Looks Like

MLConf, San Francisco, CA November 15, 2013 12

Recommender Apps

Transaction history

Customer profile

Merchant profile

Context

InputChannel

Audience

MLConf, San Francisco, CA November 15, 2013 13

Collaborative Filtering - Recommend what similar users like explicitly or implicitly.

Content based - Recommend similar items solely based on the content of items.

Hybrid- Combines the above two.

General Approaches

MLConf, San Francisco, CA November 15, 2013

Find the most relevant merchant offers for our cardmembers, with closed loop data and “real time” context.

Transactional History

LifestyleAttributes

Apr 8, 2023AXP Internal

Input to MyOffers

MLConf, San Francisco, CA November 15, 2013

BatchHadoop Environment

Contextual Information

Real TimeSolr

Offer Database

Offer Contents

CM ChannelsFulfillment

Synced Card

Merchant Reporting

Pre CalculationExpert Rules

MyOffers Ecosystem

MLConf, San Francisco, CA November 15, 2013 16

•Agile development for shorter cycle

•Platform and software challenges

•Noisy signals, e.g. taxicabs

•Cold-start issue

•Local vs. Online

Lessons Learnt

MLConf, San Francisco, CA November 15, 2013 17

Lesson Learnt – Geo-Fencing is Critical

MLConf, San Francisco, CA November 15, 2013

Current Focus is to build out an end-to-end platform and a rich experimentation layer

•Centralization of data

•Better algorithms

•Better incorporation of customer feedback

18

MLConf, San Francisco, CA November 15, 2013 19

d3.js

Custom ML Implementations

Technologies

MLConf, San Francisco, CA November 15, 2013 20

Build the next generation of:-Recommendation systems-Graph Algorithms -Machine Learning algorithms for Marketing, Fraud and a variety of problems-Data products -Experiments

We are Hiring!

MLConf, San Francisco, CA November 15, 2013 21

Please Contact us at:Abhijit Bose

VP, Data Science & Engineeringhttp://www.linkedin.com/in/abose

abhijit.bose@aexp.com

Henry YuanDirector, Data Science

http://www.linkedin.com/pub/henry-yuan/4/29b/9bbhenry.h.yuan@aexp.com

Huiming QuSr. Data Scientist, Data Science & Engineering

http://www.linkedin.com/pub/huiming-qu/4/400/b82huiming.qu@aexp.com