Prediction and Analysis of Customer Churn of Automobile Dealers Based on BI

Customer resource is an important lifeline for the survival and development of enterprises. In the context of the rapid expansion of China’s automobile sales market, with the deepening and refinement of enterprise marketing concepts, it is crucial for automobile enterprises to fully use customer consumption behavior data to mine valuable information and customize retention measures for potential lost customers. This paper takes the open dataset of auto dealer customer churn as the research object, and analyzes and models it from the perspective of business intelligence. First, data analysis (DA) and visualization methods are used to analyze and process the dataset; Then, for the problem of sample imbalance, authors use 6 sampling techniques to balance, such as SMOTE oversampling and SMOTEENN, SMOTETomek comprehensive sampling techniques to deal with it, and uses machine learning models, such as Decision Tree, Logical Regression, XGBoost, AdaBoost and other algorithms for modeling and analysis. The authors compare the models according to Recall and AUC values, and concludes that XGBoost is the optimal prediction model. The results show that the Recall and AUC of the model prediction results based on this dataset reach 0.926 and 0.842 respectively, which can help automobile enterprises effectively identify the lost customers. On the other hand, this study can also help auto enterprises customize retention measures for potential lost customers.