基于逻辑回归的客户稳定度建模

Customer Stability Modeling Based on Logistic Regression

为科学评估通信用户的忠诚度,运用专家法并通过数据抽取、转换和装载过程,从某通信企业数据库中选取120万条实验数据,利用基于互信息的属性约简算法简化冗余属性,以提高模型构建效率和质量。通过Logistic回归算法构造客户稳定度预测模型。实验结果表明,模型预测命中率提升3倍以上,能够达到实际商业应用的要求。

Aiming at the problem how to find a scientific method for assessing customer,based on data base of a western communications company,there are 1.2 million experimental data being selected through the expert methods and data extraction,transformation and loading processes.It uses attribute reduction algorithm based on mutual-information simplifies redundant attributes in order to enhance the efficiency and quality of the model construction,and through the Logistic regression algorithm the customer stability prediction model is constructed.The results show the model predicts more than 3 times to enhance hitting rate,so it can meet actual commercial applications.