Learning Item Temporal Dynamics for Predicting Buying Sessions.
IUI(2016)
摘要
Predicting whether a session is a buying session (e.g. will end with buying an item) is an ongoing research task. Drawing from recent experience in Web search and movie recommenders, we explore the effect of temporal trends and characteristics on the ability to predict buying sessions. We suggest a new approach, based on items' temporal dynamics, together with sessions' temporal aspects for predicting whether a session is going to end up with a purchase. We suggest a model for estimating the probability of a session to end with a purchase, according to the purchase history of items clicked on during the session over the past few days. The predictions can be used by recommender systems, enabling them to take relevant actions, thus improving shoppers experience as well as increasing sales for e-commerce companies. Our findings shed light on the importance of considering temporal dynamics in items recommendations in e-commerce sites. Empirical results on imbalanced e-commerce dataset with more than nine million sessions demonstrate that we achieve high Precision, Recall and ROC in predicting whether session ends up with a purchase or not.
更多查看译文
关键词
Electronic commerce, Recommender Systems, Temporal Dynamics, Machine Learning, Imbalanced Data Set
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络