A Novel Ensemble Learning Model Combined XGBoost With Deep Neural Network for Credit Scoring

Xiaowei He, Siqi Li, Xin He, Wenqiang Wang, Xiang Zhang, Bin Wang

OAI: oai:igi-global.com:299924 • DOI: 10.4018/JITR.299924

Credit scoring, aiming to distinguish potential loan defaulter, has played an important role in the financial industry. To further improve the accuracy and efficiency of classification, this paper develops an ensemble model combined extreme gradient boosting (XGBoost) and deep neural network (DNN). In the method, training set is divided into different subsets by bagging sampling at first. Then, each subset is trained as a feature extractor by DNN and the extracted features is taken as the input of XGBoost to construct the base classifier. At last, the prediction result is the average of outputs of different base classifiers. In the training verification process, three credit datasets from the UCI machine learning repository are used to evaluate the proposed model. The outcome shows that this model is superior with a significant improvement.