Huang Jiaqing, Liu Yang, Tu Miaomiao, Sohaib Osama
School of Economics, Business and Foreign Languages, Wuhan Technology and Business University, Wuhan, Hubei, China.
R&D department, Cabio Synthetic Biotechnology(Wuhan) Co., Ltd, Wuhan, Hubei, China.
PLoS One. 2025 Jul 8;20(7):e0327199. doi: 10.1371/journal.pone.0327199. eCollection 2025.
Accurate risk assessment in international trade settlement has become increasingly critical as global financial transactions grow in scale and complexity. This study proposes a hybrid model-Genetic Algorithm-optimized Fuzzy Neural Network (GA-FNN)-to enhance bank risk identification within this context. The objective is to improve the classification of bank-related risks by integrating the adaptability of fuzzy logic with the global optimization capability of genetic algorithms. The GA is used to fine-tune the structure, membership functions, and parameters of the FNN to improve predictive performance. Experiments were conducted on three public datasets: Bank Marketing, Lending Club, and German Credit. Results show that GA-FNN achieves an average classification accuracy of approximately 90% across high, medium, and low risk levels, outperforming traditional methods such as logistic regression, SVM (Support Vector Machine), and other metaheuristics like PSO (Particle Swarm Optimization) and SA (Simulated Algorithm). These findings demonstrate the model's effectiveness and practical value in dynamic international trade scenarios, offering a reliable approach for enhanced bank credit risk evaluation.
随着全球金融交易规模和复杂性的不断增加，国际贸易结算中的准确风险评估变得越来越重要。本研究提出了一种混合模型——遗传算法优化的模糊神经网络（GA-FNN），以在此背景下增强银行风险识别能力。目的是通过将模糊逻辑的适应性与遗传算法的全局优化能力相结合，改进与银行相关风险的分类。遗传算法用于微调模糊神经网络的结构、隶属函数和参数，以提高预测性能。在三个公共数据集上进行了实验：银行营销、借贷俱乐部和德国信贷。结果表明，GA-FNN在高、中、低风险水平上的平均分类准确率约为90%，优于逻辑回归、支持向量机（SVM）等传统方法以及粒子群优化（PSO）和模拟算法（SA）等其他元启发式算法。这些发现证明了该模型在动态国际贸易场景中的有效性和实用价值，为增强银行信用风险评估提供了一种可靠的方法。
