Credit Risk Assessment based on Gradient Boosting Decision Tree

Abstract

Regarding to the increasingly attention on the credit risk rating system, the tradition way to evaluate the credibility of any given individual or company using machine learning is based on methods like SVM, decision tree or MLP. In our research, a more efficient method is introduced, which is known as the Gradient Boosting Decision Tree. With proper data preprocessing and feature selection, models are compared due to their performance. Our model, Gradient Boosting Decision Tree, has been proved to be one of the best that obtain the highest accuracy (92.19%), f1 score (91.83%) and AUC value (0.97). The experiment proved that this model has the best ability of classification and generalization.