Thesis:

Presenting a method for classifying bank customers by means of combined clustering technique using black widow algorithm and K-Means

Banking is one of the key industries in the economy, and knowing the needs of customers is essential to improve services. As an unsupervised learning method, clustering divides customers based on different characteristics and helps banks to provide appropriate strategies for each group. K-means algorithm is one of the popular clustering methods, but determining the appropriate number of clusters is challenging. To solve this problem, meta-heuristic methods such as genetic algorithms and black widow spider algorithm are used. These algorithms can help to optimize and solve complex problems. In this research, the combination of black widow spider algorithm and K-Means has been used to classify the characteristics of bank customers. This research was conducted on the data of Mehr Entrepreneurship Fund (Imam Reza Institute (AS)), among the 28 features registered in the database using the proposed method, four features are the amount approved for the customer, account status, economic sector and status Employment has been extracted and selected, and the K-Means algorithm has been used to classify customers and determine whether customers are good or bad based on the four selected characteristics. The result of the proposed method was the creation of four clusters. The results obtained from the performance identification of the proposed method of total customer data showed that with 383 customers sampled using the CCR model data coverage analysis method, 311 good customers (81.1 %) and 72 bad customers (80 /18 %).

Keywords: Classification, Bank Customers, Black Widow Spider Algorithm, K-Means Algorithm