A model for validating bank customers using multilayer perceptron neural network and imperialist competitive algorithm (ICA)

Given the highly competitive nature of the banking industry, financial and credit institutions continually seek to identify the most reliable and profitable customers. They are particularly concerned about loan defaults or delays in repayment, which can negatively impact economic growth. Credit scoring models are among the most effective tools in modern banking for evaluating customer creditworthiness. These models enable banks to assess credit requests with greater accuracy and lower cost. In recent years, machine learning techniques—especially predictive classifiers-have been extensively applied to credit scoring and customer classification. This study introduces a novel hybrid model that combines a Multilayer Perceptron (MLP) neural network with the Imperialist Competitive Algorithm (ICA). In the proposed approach, ICA is employed to optimise the hyperparameters of the MLP network. The model is tested on a dataset comprising ۲,۵۷۱ real customers from Saderat Bank, categorised into default and non-default classes based on ۱۱ identified features. The results demonstrate that the proposed model achieves higher accuracy and lower prediction error in assessing customers’ credit behaviour.

کلیدواژه ها:

Credit Risk ، Credit Scoring ، Probability of Default ، Multilayer perceptron neural network ، Imperialist competitive algorithm