A novel credit scoring model based on optimized random forest

With the rapid development of the credit industry, credit scoring models has become a very important issue in the credit industry. Many credit scoring models based on machine learning have been widely used. Such as artificial neural network (ANN), rough set, support vector machine (SVM), and other innovative credit scoring models. However, in practical applications, a large amount of irrelevant and redundant features in the credit data, which leads to higher computational complexity and lower prediction accuracy. So, the face of a large number of credit data, effective feature selection method is necessary. In this paper, we propose a novel credit scoring model, called NCSM, based on feature selection and grid search to optimize random forest algorithm. The model reduces the influence of the irrelevant and redundant features and to get the higher prediction accuracy. In NCSM, the information entropy is regarded as the heuristic to select the optimal feature. Two credit data sets in UCI database are used as experimental data to demonstrate the accuracy of the NCSM. Compared with linear SVM, CART, MLP, H2O RF models, the experimental result shows that NCSM has a superior performance in improving the prediction accuracy.