| In recent years,with the rapid increase in car ownership,after-sales maintenance profits have become the main source of profit for 4S.There are many types of analysis methods to analyze the reasons for how to reduce the auto customer churn after-sales.This paper provides early warning and improvement directions for automobile after-sales marketing strategy from the perspective of constructing after-sales loss prediction model.In order to improve the classification accuracy and generalization ability of the model,this paper adopts the method of model averaging to construct a combined classification model for predicting the state of auto customer churn after-sales,evaluates the service conditions of the existing model average methods,and makes an empirical analysis after reasonably processing the candidate classification models according to the conditions.The model averaging method used in this article mainly relies on maximum likelihood estimation and least squares method to obtain the weight distribution of sub-models,and Kullback-Leibler divergence is also used many times as a measure to judge the goodness of fit between real data and model estimation.In order to obtain valid data samples from the original data set with high dimensionality,large sample size,and low data quality,this paper uses multiple detection methods and a variety of statistical analysis tools to screen,correct,and eliminate noise and invalid features of the original data set and does all the other necessary data cleaning.The empirical results show that the classification model built on this data set,the classification performance and classification effect meet the demand.By using the model averaging method,the experimental results show that the classification performance of the combined prediction model constructed by model averaging method has been improved,which also shows that the model averaging method can promotes the generalization ability of the classification model and avoid over fitting of a single classifier.At the same time,the constructed classification model reveals the main factors that affect customer churn after sales.The classification model meeting business needs can play an early warning and reminder for the classification of after-sales customer churn. |
