基于logistic回归的睡眠预测模型

发布时间：2018-06-11 21:16

本文选题：逻辑回归 + 睡眠预测模型　；参考：《云南大学》2015年硕士论文

【摘要】：某第三方支付平台存在大量睡眠用户。因为不产生任何效益所以可以认为睡眠用户就是客户流失。这对于运营部门而言是较大的浪费。因此,从维系公司运营角度考虑,准确找出哪些用户进入睡眠,那么在市场部挽留措施下则能有效提升用户的价值。本文以某支付平台用户作为研究对象。从用户行为分析的角度和现有数据基础上确定建模变量。在单变量分析上,运用elogit散点图分析对连续变量进行变量变换及判断变量对模型的影响效应,运用频率分析对类别变量进行分析。利用二元logistic模型建立移动支付平台用户睡眠预测模型,结合实际数据进行实证研究。结果：最后一次交易后余额,最近三月消费次数,最近三个月消费笔数占比,最近三个月交易金额占个月总交易额百分比,最近三个月转账笔数占比,实名等级,最近三月转账金额占比与最近4-6个月转账金额占比变化比例对用户睡眠具有统计学意义。其中最后一次交易后余额,最近三月消费次数,最近三个月消费笔数占比,最近三个月交易金额占六个月总交易额百分比影响最为显著。
[Abstract]:A large number of sleep users exist on a third-party payment platform. You can think of sleep users as customer churns because you don't have any benefits. This is a big waste for the operations department. Therefore, from the point of view of maintaining the operation of the company, finding out exactly which users are sleeping can effectively enhance the value of the users under the retention measures of the marketing department. In this paper, the user of a payment platform as the research object. The modeling variables are determined from the point of view of user behavior analysis and existing data. In univariate analysis, elogit scatter plot analysis is used to analyze the variable transformation of continuous variables and the effect of judging variables on the model, and frequency analysis is used to analyze the class variables. The sleep prediction model of mobile payment platform is established by using the binary logistic model, and the empirical research is carried out based on the actual data. Results: the balance after the last transaction, the number of times consumed in the last three months, the number of pens consumed in the last three months, the transaction amount in the last three months as a percentage of the total monthly transaction volume, the number of transfer pens in the last three months as a percentage of the real name grade, The proportion of transfer amount in recent March and in the last 4-6 months is statistically significant to the sleep of users. The balance after the last transaction, the number of times spent in the last three months, the number of pens consumed in the last three months, and the percentage of the transaction amount in the last three months to the total trading volume in six months were the most significant.
【学位授予单位】：云南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O212.1

【参考文献】