With the rapid growth of data in today’s world, leveraging data science to develop optimal solutions has become essential. One of the key challenges in the banking industry is ensuring the repayment of granted loans and the efficient allocation of financial resources. Therefore, assessing and predicting customer credit levels can serve as a reliable and profitable indicator for banks. In this study, based on financial transaction data from customers of Mellat Credit Institution, the current credit level was analyzed using the RFM model combined with clustering techniques. In the proposed approach, the future credit level of customers was also predicted using a Random Forest machine learning algorithm built upon the results of model-based clustering. Compared with traditional static scoring methods, this approach improves prediction accuracy and enables more confident credit decisions. Using the Bayesian Information Criterion (BIC) to identify the optimal model and cluster count, five distinct customer clusters were detected through the RFM framework. Furthermore, a Random Forest model achieved an average accuracy of 72%, recall of 75%, and an F1-score of 91% for predicting customers’ future credit levels. By incorporating modern financial technologies, this research presents a data-driven framework for intelligent credit scoring in Islamic banking, contributing to enhanced predictive performance and reduced credit risk in loan allocation processes.