基于改进支持向量机的产品服务系统客户流失预测模型

摘要

To assist the designer accurately to provide the effective solution of product service system for customers,according to the customer churn prediction problem,a kind method of customer churn prediction combining an improved particle swarm algorithm with support vector machine (IPSO-SVM) was proposed. The method includes constructing the customer churn model of product service system and the IPSO-SVM al-gorithm model. Firstly,the particle position is used to represent the parameters of SVM, and the position and velocity of particle swarm are initialized based on Sobol sequence. Finally, through compared with BPNN, SVM, PSO-SVM, the feasibility and validity of the introduced method was verified by appling in Customer churn prediction model of product service system for CNC Machine Tools.%为协助设计师能精准地为客户提供有效的产品服务系统整体解决方案,针对其客户流失预测问题,提出了一种改进粒子群算法与支持向量机相结合的客户流失预测方法(IPSO-SVM).该方法包括构建了产品服务系统客户流失模型及IPSO-SVM算法模型.首先,IPSO-SVM算法采用粒子位置表示支持向量机的参数,并基于Sobol序列对粒子群位置与速度初始化,然后位置更新时引入动态自适应非线性惯性权重的方法.最后,以某高档数控机床公司客户流失状态为案例,通过与BPNN、SVM、PSO-SVM进行比较,验证所提方法在该数控机床产品服务系统客户流失模型中的有效性与可行性.