Financial customer classification by combined model

Abstract

This paper explores the pros and cons of different algorithm models on the same selection problem, and then uses the combined prediction theory to obtain a new combined prediction model to explore its prediction accuracy. The actual problem to be solved is to help financial institutions to scientifically classify customers who choose financial products. We select the bank data set in the UCI database, which is derived from the survey data of a customer conducted by a financial institution in Portugal for a wealth management product. Decision tree C5.0 algorithm, naive Bayes classification algorithm and binary logit model are individually used to carry out a single model of empirical research on financial product customer classification. Through the empirical analysis of the five combination models, it is concluded that in the model that uses the least squares weighting method to determine the weight, the weight appears negative, which does not conform to the actual situation. The model that is based on the least squares weighting method and the model that is based on the simple weighting method are excluded. In contrast, the arithmetic mean weighted model is better than the reciprocal variance weighted model and the reciprocal mean square model. The accuracy reaches 89.91%, which is 0.43% higher than the accuracy of a single model. It can be concluded that the model that is based on the arithmetic average weighting is a better combination forecasting model.