摘要
首先对用户数据进行特征分析,变量选择,然后又采集了大量与手机性能相关的数据来扩充数据集,最后利用现代数据挖掘手段对用户的换机行为进行预测,讨论并比较了各种方法对换机预测的准确性.通过对用户数据集进行测试实验,表明变量选择与补充能够有效地提高移动用户换机的预测结果,并且Xgboost方法在各种分析工具中的表现更为优越.
This paper, first of all, makes an analysis of the characteristics of user data, variable selectioa, and then collects a large number of new data to expand data sets. At last, this paper makes use of modern data mining methods to predict mobile users for updating terminal behavior, discusses and compares the various methods＇ accuracy. Based on user data set for testing experiment, it shows that variable selection and adding variables can effectively improve the accuracy of prediction, and Xgboost is more superior performance in various analysis tools.
出处
《数学的实践与认识》
北大核心
2017年第16期71-80,共10页
Mathematics in Practice and Theory
基金
国家自然科学基金
高维数据变量间非线性交互作用的研究(11571009)
关键词
添加变量
变量选择
换机预测
Xgboost
adding variables
variable selection
users for updating terminal prediction
Xgboost
