基于Bagging策略的XGBoost算法在商品购买预测中的应用  

The Application of XGBoost Algorithm Based on Bagging Strategy in the Pprediction of Commodity Purchase

作　　者：谢冬青[1] 周成骥 

机构地区：[1]广州大学计算机科学与教育软件学院,广东广州510006

出　　处：《现代信息科技》2017年第6期80-82,共3页Modern Information Technology

摘　　要：随着电子商务的发展,越来越多的消费者在线上购买商品。消费者在登录购物网站时,往往会留下大量的历史行为数据。如何通过消费者的行为记录,预测消费者未来的购买行为,并以此为依据,设计推荐系统,成为越来越多的学者研究的重点。本文将XGBoost算法应用到商品购买预测中,并使用Bagging集成学习方法对单一算法进行改进,以提高预测的准确性。最后通过实验证明,采用Bagging集成学习方法的XGBoost算法模型整体效果上明显优于单一算法的模型。With the development of e-commerce,more and more consumers buy goods online.When consumers log on to the shopping sites,they tend to leave a lot of historical behavior data.How to predict consumers＇ future purchase behavior through consumer behavior records,and based on this,design recommendation system has become the focus of more and more scholars.In this paper,the XGBoost algorithm is applied to the prediction of commodity purchase,and the Bagging integrated learning method is used to improve the single algorithm in order to improve the accuracy of the prediction.Finally,the experiment shows that the overall effect of the XGBoost algorithm model using Bagging integrated learning method is obviously better than the single algorithm model.

关 键 词：数据挖掘 购买行为 XGBoost 集成学习 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]