An improved credit card users default prediction model based on RIPPER

With the vigorous development of the financial sector, financial risks are showing a tendency of diversification, which the customer credit risk of commercial Banks in particular. As a result, the customer credit risk is generally taken into account by financial institutions, credit evaluating model emerges as the times require. At present, many researches have concentrated on enhancing the precision of the model, ignoring the interpretability, which makes it difficult to apply in industry; Compared to Precision, the study of interpretable model is relatively small, moreover, due to imbalanced data sets, the accuracy of existing models is not high. Therefore, this study proposes an improved model based on RIPPER algorithm. For the problem of credit card data sets imbalanced distribution, using the synthetic minority class sampling algorithm to equalization samples, then taking advantages of rules generated by RIPPER algorithm to forecast default credit card users. To test performances of the model, we use the real Taiwan credit card customer datasets for the empirical research. From perspectives of the model accuracy and interpretability, comparing the proposed SRIPPER model with the existing mainstream models, results of experiments indicate that the proposed model obtains good results. This study demonstrates that the proposed credit card user default prediction model SRIPPER has certain practical application value.