Abstract

Credit scoring has grown in importance and has been thoroughly researched by banks and financial institutions. The amount of redundant and irrelevant features present in credit scoring datasets, however, reduces the classification accuracy. As a result, employing effective feature selection methods has become essential. In this study, a hybrid feature selection approach that combines the backpropagation neural network (BPNN) classifier and the pigeon optimization algorithm (POA) is suggested. With hybridization, the POA works to choose characteristic subgroups through the feature selection (FS) process, and the BPNN then assesses the chosen subsets using a fitness function. The experiment findings show that the suggested hybrid technique outperforms other competing approaches in terms of evaluation criteria, according to three benchmark credit scoring datasets.