基于数据挖掘的证券营业部客户流失分析

Analysis of Customers Churn and Data Mining from A Securities Business Department

针对营业部客户流失的影响因素数据给出了预测模型。根据新佣金政策下的证券营业部客户一年来的交易数据分析,提出了影响客户流失的各种特征因素的假设。进一步建立三个模型(逻辑回归、决策树和径向基神经网络),并从错分率和预测能力(提升值)两方面的模型性能评判标准来选出最佳模型,根据该模型找出最可能影响客户流失的各种预测因子,并进行深入分析,最后提出相应的对策。

The paper builds a forecast model with factors to affect customers＇ churn. It presents some hyphothesises by these factors from the data of a securities business department, and builds three models （logistics, decision and nural network） and selects the best one from the above models by both disclassification and forecast ability. Moreover, it finds some factors which affect customers chum with maximum probability and presents some policies to prevent customers from chum.