A Study on a Comparison of the Classification Model Predictive Results A Case Study Investment Risk of Customers in Bank A

• Main Authors: Shih-Hung Chen, 陳世鴻
• Other Authors: Te-Hsin Liang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/9myaq3

碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 102 === Currently data mining and machine learning classification technique is widely used in various fields, and which classification method is the best？To find the most effective way to predict the best result is the most important thing we concerned about. Therefore, evaluating the result of the classification method is worthy to study in practice. In this study, we use the data of bank A to discuss customers of bank who are under different factors will choose which investment risk preferences. Therefore, this study will use three different methods to select variables. So we use Multivariate Adaptive Regression Spline, Classification and Regression Tree, Random Forest, Back Propagation Neural Network, Support Vector Machine and Multinomial Logistic Regression to build models. Comparing six kinds of classification methods to find the best classification model. The results are shown in Multivariate Logistic Regression model was optimal. It gets higher, more balanced and stable classification accuracy rate. It has better ability to explain the model.