摘要
为了提高电子商务中的用户购买行为预测的效果,提出数据平衡与模型融合的方法来对用户购买行为进行预测。在对用户行为数据提取特征时发现数据样本类别存在严重不平衡的情况,针对这一问题运用改进的欠采样平衡方法处理用户行为数据,运用基于极端梯度提升(XGBoost)算法的融合模型对用户购买行为进行预测。以京东商城的交易数据作为实验数据集,通过与单一预测模型的对比实验证明了在预测精度和泛化能力方面融合模型相较于单一预测模型的表现都更好。
In order to improve the effect of prediction of user purchase behavior in e-commerce,this paper proposes a method of data balance and model fusion to predict users’purchase behavior.When extracting features from user behavior data,it is found that there is a serious imbalance in the categories of data samples.To address this problem,an improved under-sampling data balance method was used to process user behavior data,and a fusion model based on the XGBoost algorithm was used to predict user s’purchase behavior.Taking the transaction data of Jingdong Mall as the experimental data set,the comparative experiment with the single prediction model proves that the fusion model performs better than the single prediction model in terms of prediction accuracy and generalization ability.
作者
李伊林
段海龙
林振荣
Li Yilin;Duan Hailong;Lin Zhenrong(Jiangxi Province Institute of Water Sciences,Nanchang 330029,Jiangxi,China;School of Information Engineering,Nanchang University,Nanchang 330031,Jiangxi,China)
出处
《计算机应用与软件》
北大核心
2022年第9期50-55,86,共7页
Computer Applications and Software
基金
江西省自然科学基金项目(20161BAB212040)。
关键词
购买预测
数据平衡
极端梯度提升
融合模型
Purchase forecast
Data balance
Extreme gradient boosting
Fusion model
