客户购买行为建模分析预测

Customer Purchase Behavior Modeling, Analysis and Prediction

基于阿里巴巴天池大数据比赛的真实客户购买记录数据,通过分析客户以往的购买记录,预测未来客户对哪些商品会有行为。采用先召回部分商品、再进行模型预测排序的策略来提高预测效率。在召回过程中,对传统的根据商品类别召回商品的方法加以改进,加入对用户行为时间顺序的考虑,排序过程中采用XGBoost、LightGBM、CatBoost等boosting算法进行排序,从而有效预测未来用户会对哪些商品有所行为。

Based on the real customer purchase record data of Alibaba Tianchi big data competition and analyzing the customer’s previous purchase records,we can predict which items that customers will act in the future.In order to improve the prediction efficiency,the strategy which re⁃calls some items first and then ranks items by model is adopted.In the process of recalling,this paper improves the traditional method of re⁃calling items according to category by considering the time sequence of customer behavior.In the process of ranking,this paper uses some boosting algorithms like XGBoost,LightGBM,CatBoost to rank so as to predict which items that customer will act in the future.