基于铁路货运价值分类的客户流失预测研究

Research on railway freight customer churn prediction based on customer value segmentation

结合铁路货运行业的特征,获取其他货运市场数据,从客户发货情况、客户服务情况和运输市场情况3方面建立客户流失识别方法。根据铁路货运特征提出基于RFM的货运客户价值分类模型KFA和货运客户价值的计算方法。运用k-means聚类算法对货运客户进行分类,并利用支持向量机(SVM)建立各类货运客户的流失预测模型。制定评估标准来验证预测模型的预测效果。仿真结果显示,KFA客户分类模型具有较好的分类效果,按照不同客户分类建立支持向量机客户流失预测模型具有较强的预测能力,且对于不同观察窗口的数据分析结果差异性较小,说明模型具有较强的泛化能力,并且相比于全局预测,对于高价值客户更具有准确性,具有较高的实际应用价值。

By combining with the characteristics of the railway freight transportation and accessing other freight market data,this paper established a customer churn recognition method from the perspectives of customer delivery performance,quality of customer service and transport market situation.According to the characteristics of railway freight,the freight customer value segmentation KAF model was put forward on the basis of RFM model.The prediction model for all kinds of customer value segmentation was established by using support vector machine(SOM)besides,the evaluation criteria for verifying the prediction effect of the model were also proposed.The simulation results show that KFA customer segmentation model has better segmentation effect,and support vector machine churn prediction model has strong predictive ability according to different customer segmentation.The analysis results have small differences for different watch windows,which shows that the model has strong generalization ability.Compared with global prediction,the prediction for high value customers is more accurate and the proposed models have high practical value.