A personal credit scoring model based on machine learning method

摘要

#$%^&*AU2018101523A420181115.pdf#####ABSTRACT The present invention, mainly applied to the Chinese financial industry, is concerned with optimization and processing of personal credit scoring. As for the data preprocessing, attribute mean is used to fill in the missing values, moreover, training samples containing the outliers are ignored and normalizing features uses Min-Max Normalization. The raw data contains a large number of redundant features, so the degree of importance of the feature is calculated by the random forest, and the top 90 features are selected as the final sample feature according to the results calculated by the random forest. In the model selection, because of the convergence and promotion ability of SVM, the personal credit scoring model based on the SVM has strong practicability.Data Colllection Missing value Processing Data Preprocessing Delete Outliers Data Normalization Processing Sample Selectilon No Model TriinJg Support Vector ROC curve Machine Precision Model Evaluation Recall Meet the Requirements? Finish Figure 1 1