A Tool to Predict Payment Default in Financial Institutions

Loans are financing services for clients of a bank and are one of the main activities in a financial institution since they are the means through which they make money. When a customer misses one or more payments cause grave problems at the bank at the point of crash. The bank loan manager to decide decides whether to approve or not the loan application using the client’s financial and personal information. This decision always has associated risks. Currently, financial institutions, to reduce the risks associated with loan approval and take advantage of the large repositories of historical data from their clients, are using machine learning algorithms to identify if a client will comply with the loan payment. That information helps managers in their decision-making process. This paper presents the development of an application to support the process of authorizing or not a bank loan in the Acción Imbaburapak Savings and credit cooperative; to choose the model to use in the application, select after training three predictive methods. The analytical process followed the phases proposed by the KDD methodology. Three supervised classification methods were selected: logistic regression, decision trees, and neural networks. Since the neural network showed the best results during the evaluation, we chose this to build the application.