摘要
为提高对用户购买意向预测的准确率,提出了一种基于堆叠法集成学习的用户购买行为预测模型。利用模型融合技术,将逻辑回归、决策树和XGBoost模型作为基学习器输入,再以随机森林模型作为次学习器进行堆叠,从而形成一种组合模型。针对电商提供的线上用户数据集,首先利用滑窗技术提取用于预测用户购买行为的特征,然后分别使用逻辑回归、决策树、XGBoost和集成学习组合模型预测用户购买意向的准确性。结果表明,组合模型的准确性明显优于其他算法。
In order to improve the accuracy of user’s purchase intention prediction,a prediction model of user’s purchase behavior is proposed based on stacking integrated learning.The model fusion technology is used,and logistic regression,decision tree and XGBoostare taken as the input of the base learner;and then the random forest model is stacked as the secondary learner to form a combination model.According to the online user dataset provided by e-commerce,the sliding window technology is used to extract features to predict user purchase behavior,and then logistic regression,decision tree,XGBoostand integrated learning combination model are combined to predict the accuracy of user purchase intention.The results show that the accuracy of the combined model is better than other algorithms.
作者
谢日敏
方雨桐
XIE Rimin;FANG Yutong(College of Information Engineering,Fujian Business University,Fuzhou 350012,China;College of Mathematics and Computer Science,Fuzhou University,Fuzhou 350108,China)
出处
《重庆科技学院学报（自然科学版）》
CAS
2021年第3期70-73,共4页
Journal of Chongqing University of Science and Technology：Natural Sciences Edition
基金
福建省自然科学基金面上项目“多波束卫星通信系统的多目标动态信道分配问题研究”(2020J01323)
福建省中青年教师教育科研项目“基于深度学习的学生课堂学习状态研究”(JAT170694)。
关键词
集成学习
堆叠法
用户购买意向预测
决策树
XGBoost
integrated learning
stacking method
forecast of user’s purchase intention
decision tree
XGBoost
