基于多算法融合的移动通信客户流失预测模型

Customer Churn Prediction Model of Mobile Communication Based on Multi-algorithm Fusion

针对移动通信行业中客户不断流失的现状,提出了一种优于传统单一算法模型预测的组合模型。该组合模型的元模型分别为决策树模型、Logistic回归模型和BP神经网络模型,该模型综合了各个元模型的优势。通过构造拉格朗日函数的方式来确定每个元模型的最优权重,使组合后的预测模型达到最优的预测效果,并在某移动通信公司提供的数据仓库中随机选取足够数量的流失客户作为数据集进行实验。实验结果表明,该模型在预测的正确率上比每一个元模型均有明显的提高。该方法有很好的预测效果,能够帮助移动通信公司找出即将离网的客户,对其制定相应的业务来维护自身商业利益。该方法的局限在于仅考虑了各个元模型间线性组合的情况。

Aiming at the present situation of the continuous loss of customers in the mobile communication industry,we propose a combinatorial model which is superior to the traditional single algorithm model,of which the meta-model are the decision tree model,the logistic regression model and the BP neural network model.It combines the advantages of each meta-model.The optimal weight of each meta-model is determined by constructing the Lagrange function,and the combined prediction model is used to achieve the optimal prediction effect.In a data warehouse provided by a mobile communication company,a large number of lost customers are selected randomly as the data sets for experiment which shows that the model has a significant improvement in the accuracy of each model.The method has a great predictive effect,and can help mobile communication companies to find out the off-line customers,to develop their own business to maintain their own business interests.The limitation of this method is that only the linear combination of each meta-model is taken into account.