Prediction of Customer Purchase Probablity for Online Recommendation Systems

Abstract

Tracking of online customers’ behavior is essential part in online recommendation systems. Specifically, for online recommendation at online stores, it is necessary to understanding customers’ purpose through their real-time activity data. This study suggests state probability methods that predict customers’ purchase probability using their clickstream data. Also, it verifies usefulness of the proposed method by using real clickstream data of an online book store. From experimental results, state probability models show better performance. Also, 2-state model with weight show best performance among proposed methods