فهرست نشریات علمی دسترسی آزاد ایران

Nowadays, how organizations, especially banks, interact with customers through Customer Relationship Management (CRM) has significantly changed. The present study aims to explain a customer classification model using fuzzy Bayesian recommender systems. This is an applied and exploratory research in which the information of 98,604 customers of one of the Iranian banks, randomly selected from the banks active in the country, was analyzed using data mining, fuzzy, and nonlinear Bayesian Model Averaging (BMA). This research, 22 customer-related indicators were entered into nonlinear Bayesian models (TVP-DMS,  TVP-DMA,  and BMA). According to the rate of error, BMA had the most excellent accuracy. The results identified the variables "account balance", "total deposit balance", "total current facility balance", and "volume of financial transactions" as non-fragile variables. Next, the results indicated the higher accuracy of the C-MEANS approach than the K-MEANS approach. Then, 16 clusters were identified using the C-MEANS approach, and their characteristics were analyzed. As a result, the selected variables of the BMA approach were used to evaluate neural network and meta-heuristic models. The BMA model had the most excellent accuracy according to the error rate. After determining the model, four main variables were specified as non-fragile variables: account balance, total deposit balance, total current facility balance, and the volume of financial transactions. This section extracted an optimal neural network model for each identified cluster based on the minimum prediction error. The results revealed a good consistency between the neural network approaches and customer classification using the data mining approach in the sign and effect size. This research focuses on designing an intelligent hybrid model combining neural networks and nonlinear Bayesian modelling for customer classification in banks. The model improves prediction accuracy using Type-3  fuzzy clustering and enables the simulation of customer behaviour patterns. Practical implications include enhanced decision-making, CRM, risk reduction, and service personalization.