Abstarct: Abstract The banking industry is responsible for long-term and short-term financing of economic activities and contributes to the stability of financial systems of countries. Also, the bank is the main institution active in the field of credit provision, and granting loans and facilities is an important part of the financing operations of any bank, but the possibility of not repaying loans and facilities on time causes credit risk in banks, and neglecting this issue can lead to unfavorable results in the performance of banks. Credit risk management is an important part of a comprehensive risk management method and a basic condition for the long-term success of any bank. Therefore, banks need a system for credit scoring of customers to reduce credit risk by recognizing and identifying customers' behavior. Considering the importance of this issue, the present study examines the factors affecting the behavior of bank customers. In this research, real data of one of the state-owned banks has been used, and machine learning, data mining, and graph mining methods have been used to achieve reliable and high-quality results. At first, by using three methods of data correlation calculation, implementation of clustering algorithm and graph embedding, feature selection and extraction were done, then the classification algorithm was implemented on the data set obtained from each of the methods. The results showed that the combined use of clustering and classification algorithms can increase the accuracy and quality of the results. In addition, using the output of the graph embedding algorithm increases the classification accuracy by more than 5% on average compared to other methods, due to the consideration of the nodes and their neighbors, and makes it possible to identify the factors affecting the customer's behavior with higher accuracy.