چكيده به لاتين
One of the problems faced by the banking system of various countries, including Iran, is the increasing growth of overdue receivables in the banking network, which indicates a decrease in the quality of banking network assets and, consequently, possible financial instability in the future, so The correct pricing of facilities, the correct allocation of risk costs to different borrowers and the correct estimation of the risk of the borrowers have become very important. To control this trend in recent decades, many banks have resorted to credit methods. In the accreditation process, the necessary decisions are made to grant facilities using methods based on financial records, according to which customers are divided into credit groups. The purpose of this study is to use data mining technique to predict the behavior of bank customers that can be categorized well based on indicators and parameters influencing the selection of credit applicants and reduce the possibility of non-repayment of facilities granted. In this research, which is of practical type, data mining techniques based on CRISP-DM process and SpssModeler18 software have been used. In the first stage, by considering different scenarios for the target field and executing different classification algorithms, an attempt was made to identify a model that has a higher efficiency in terms of accuracy and precision. After identifying the model, in order to increase the accuracy of the model, ten Boosting combined classification methods were used and finally, the cross-validation method was used to evaluate the model and cover the data-driven nature of the selected model.
