Influencing Factor Analysis for Automobile Customers Based on One-Way ANOVA and Decision Tree

Abstract

Customer-related companies are always doing researches on the customer behaviors in order to promote the profitability, and one of the most representative corporations is the automobile companies. For long time they are studying on the customer segments and using different methods to treat them. Therefore, this paper studies the influencing factors of the customer spending level and the classification method to determine different customer segments and give predictions for future customers. Firstly, method of One-Way ANOVA is used to determine the significance of influence on each factor, and the result shows that attributes including Age, Gender, Graduated, Married, Work Experience and Family Size have a determining impact on the customer spending level. Then, based on all these variables included, a decision tree model is constructed using CART method with Gini Index and Pruning for cost complexity method. The depth of the tree is 5 and the overall accuracy reaches 81.6%. Compared with other decision tree model, CART tree gives a high accuracy level with relatively simple tree structure. These results could assist automobile companies to make appropriate customer segments and develop corresponding strategies, which could be further expanded to other customer-targeted fields.