Proceedings of the 2nd International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2023, July 7–9, 2023, Chongqing, China

Research Article

Prediction of Bank Product Subscription Behavior Based on Random Forest Algorithm

Download36 downloads
  • @INPROCEEDINGS{10.4108/eai.7-7-2023.2338033,
        author={Tao  WEI and Linying  ZOU and Yanze  SUN and Shengfa  ZHAO},
        title={Prediction of Bank Product Subscription Behavior Based on Random Forest Algorithm},
        proceedings={Proceedings of the 2nd International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2023, July 7--9, 2023, Chongqing, China},
        publisher={EAI},
        proceedings_a={FFIT},
        year={2023},
        month={10},
        keywords={random forest machine learning bank product subscription big data},
        doi={10.4108/eai.7-7-2023.2338033}
    }
    
  • Tao WEI
    Linying ZOU
    Yanze SUN
    Shengfa ZHAO
    Year: 2023
    Prediction of Bank Product Subscription Behavior Based on Random Forest Algorithm
    FFIT
    EAI
    DOI: 10.4108/eai.7-7-2023.2338033
Tao WEI1,*, Linying ZOU1, Yanze SUN1, Shengfa ZHAO2
  • 1: Fuzhou University
  • 2: Qinghai University for Nationalities
*Contact email: 534011694@qq.com

Abstract

Based on the background of bank customers' purchase of products, this study uses random forest to predict whether customers buy bank products and discusses the advantages of random forest compared with other commonly used machine classification models. By selecting the product data set of Aliyun Tianchi Bank customers and using python for data processing, this study optimizes the nestimators, maxfeatures and other important parameters in random forest, so as to achieve the optimal effect of random forest classification. Meanwhile, by comparing KNN, logistic regression, support vector machine, Single decision tree model, the confusion matrix and ROC curve were used to evaluate the model performance. The final experimental results show that the random forest model with optimized parameters has better classification effect than other binary classification models.