Utilization of Data Mining to Predict Non-Performing Loan

In the banking industry, the existence of problem loans is inevitable. NPL (Non-Performing Loan) will certainly have an impact on the reduction in the capital of a bank. One good step in reducing the risk of credit default or the emergence of non-performing loans is to take proper care of debtors who begin to experience payment constraints. The main obstacle experienced in bank management, especially in the credit sector, is being unable to identify or detect potential debtors early due to a large amount of data and manual processing. In this study, the debtor payment history is presented as data to predict the existence of problem loans. History payment can be used to predict bad loans. The technic of data mining in this experiment is a new method. The results of research conducted using Nave Bayes, Decision Tree, K-NN, Rule Induction, Logistic Regression, Random Forest, Generalized Linear Model, and Gradient Boosted Trees as a comparison then choose the method that has the highest accuracy to be implemented in making additional modules on the core banking system. Random Forest is the model that has the highest accuracy of 96.55%.