基于粒子群优化BP神经网络的汽车4S店客户流失预警

Customer Churn Warning for Automobile 4S Stores based on Particle Swarm Optimization BP Neural Network

客户流失预警作为防止汽车4S店客户流失的重要手段,不仅为当代车企提供了有效的经济效益保证,也为车企对未来决策带来了新的研究依据。为建立汽车4S店客户流失预警分级标准,该文从客户基本信息、车龄、车辆销售价格、贷款金额、维修保养次数、维修保养时间等29个指标着手,基于粒子群优化BP神经网络算法,建立汽车4S店客户流失预警分级标准模型。该模型首先预测出客户流失概率,然后根据值为0-1之间的概率大小分为1-5共5个等级,其中1表流失可能性很小,5表示流失可能性很大。最终得到测试集客户流失预警从1到5等级的比例分别为71.39%、3.75%、3.50%、5.86%和15.50%。同时,通过训练集中有78.65%的客户未流失作为先验概率,判定预测概率小于等于先验概率为客户未流失,大于先验概率为客户流失,得到该模型总体的准确率为91.71%。

As an important means to prevent the loss of customers in automobile 4S stores,customer churn early warning not only provides an effective economic benefit guarantee for contemporary car companies,but also brings a new research basis for car companies to make future decisions.In order to establish the grading standard of customer churn early warning in automobile 4S stores,this paper starts from 29 indicators such as customer basic information,vehicle age,vehicle sales price,loan amount,maintenance times,and maintenance time,and establishes a standard model for customer churn early warning in automobile 4S stores based on the particle swarm optimization BP neural network algorithm.The model first predicts the probability of customer churn,and then divides it into 5 levels from 1 to 5 according to the probability size of the value between 0 and 1,where 1 indicates that the probability of churn is very small,and 5 indicates that the probability of churn is very large.In the end,71.39%,3.75%,3.50%,5.86%and 15.50%of the test set customer churn warnings were obtained,respectively.At the same time,78.65%of the customers in the training set are not churned as the prior probability,and the prediction probability is less than or equal to the prior probability of customer churn,and the overall accuracy of the model is 91.71%.