AUTOMATED ADAPTIVE MODELING IN BIG DATA RECOMMENDER SYSTEMS: THE CASE OF MOBILE AD PLACEMENT

摘要

Recommender systems (ReCo's) have become a familiar artifact in cyberspace as a vehicle for increasing revenues while deepening customer loyalty and satisfaction. Typical Reco's are used as cross-sell instruments to encourage existing customers to buy products and services related to previous purchases. However, this form of customer targeting is not the only one which ReCo's can perform. We show how to extend the conventional functionality of ReCo's to a very dynamic form of customer targeting as manifested in Mobile advertisement placement. This type of ReCo must be able to coordinate and access multiple very large databases, perform automated model generation to build large-scale logistic regressions, and implement adaptive modeling in the form of model balancing to reflect current user behavior during marketing campaigns, all in near real-time or extreme real-time. ReCo requirements for predictive real-time analytics necessitate the implementation of current "big data" software and hardware. We discuss the anatomy of such a recommender system designed for placement of Mobile advertising.