Publication

Abstract : In the present era, Credit Risk Evaluation (CRE) holds extreme technical significance and real-world use in banking sectors. Accurate and efficient credit risk assessment models are instantly essential because of the problems where data processing consumes more time to build and a huge number of data mining stages. In the field of CRE and data storage, blockchain technology along with deep learning models have been widely suggested for ensuring the privacy and security to the customer’s credit data. However, the traditional approaches are not very interpretable. To this concern, this research introduces a Multi-layer Residual coordinate Boosted Sooty tern Attention Network (MRBSAN) with consensus-based blockchain technology for secure storage of evaluated customer’s credit data in the banking sector. The introduced approach includes effective preprocessing of data, optimal selection of features by enhancing with pheromone concentration, evaluation of credit risk and secure storage of credit data by utilizing focal loss to achieve a remarkable output. The findings demonstrate that the proposed approach effectively assesses the customer’s credit risk and stores the credit data securely. The proposed model has an average accuracy of 97.2%, precision of 98.6%, recall of 98.6%, and an F1-score of 98% using three datasets. These results surpass the performance of existing approaches.