A reinforcement learning and predictive analytics approach for enhancing credit assessment in manufacturing
Razaque A. Beishenaly A. Kalpeyeva Z. Uskenbayeva R. Nikolaevna M.A.
June 2025Elsevier Inc.
Decision Analytics Journal
2025#15
The fundamental issue with a credit system for manufacturers and importers of commodities is inefficient credit assessment. Traditional techniques frequently produce inaccurate risk assessments and credit scores, resulting in financial losses for lenders, missing business growth possibilities, and less favorable client conditions. To overcome this issue, a comprehensive credit assessment scoring system should be implemented to increase importers’ confidence. The article proposes a predictive-based reinforcement learning (PRL) model to help manufacturers and importers acquire more accurate and dependable credit scores while avoiding default risk. Furthermore, the proposed PRL model enhances decision-making, system efficiency, and risk-tolerant financial conditions. To attain these cutting-edge objectives, the proposed PRL model combines three algorithms. Algorithm 1 collects and aggregates data to indicate areas for improvement if credit scoring is poor. Algorithm 2 uses reinforcement learning to validate and enhance bank scores. Algorithm 3 focuses on predictive modeling for bank scoring, ensuring that the credit decision-making system is operational and constantly improving. Furthermore, reinforcement learning leverages the features from local interpretable model-agnostic explanations (LIME) and shapely additive explanations (SHAP) to generate locally reliable explanations and attribute the contribution of each feature for determining the output of the model. The Python platform tests the proposed PRL to achieve the objectives. Based on the results, The PRL model markedly enhances credit assessment precision, achieving an accuracy of over 99.5%, which outstrips current methodologies such OCLA (96.12%), PSML (84.12%), and EMPCC (91.67%). Furthermore, the PRL model augments leverage ratios, rising from 2.75% in 2015 to 3.36% in 2024.5, and increases accounts receivable turnover from 4.38% in 2015 to 7.4% in 2024.5, surpassing alternative credit evaluation methodologies. This research highlights the novelty of combining predictive analytics and reinforcement learning to revolutionize credit assessment, providing a scalable and reliable solution for manufacturers and importers. The findings establish the PRL model as a transformative approach for creating risk-tolerant and efficient financial environments.
Financial analytics , Intelligent credit systems , Machine learning , Manufacturing , Predictive analytics , Reinforcement learning
Text of the article Перейти на текст статьи
Department of Software Engineering & AI, School of Computing, Gachon University, Seongnam-si, South Korea
Department of Software Engineering, Institute of Automation and Information Technologies, Satbayev University, Kazakhstan
Department of Computer Engineering, International Information Technology University (IITU), Almaty, Kazakhstan
Department of Software Engineering & AI
Department of Software Engineering
Department of Computer Engineering
10 лет помогаем публиковать статьи Международный издатель
Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026