A novel hybrid deep long short-term memory – multilayer perception model with feature fusion for loan risk prediction

Loan risk prediction is a crucial task for financial institutions to assess the likelihood of a borrower defaulting on their loan. This research presents a novel approach for loan risk prediction using a hybrid Deep Long Short-Term Memory (LSTM) - Multilayer Perception (MLP) architecture by feature fusion. Feature fusion is a crucial technique in loan prediction that involves combining multiple features or variables to create a new, more informative feature. The proposed architecture combines the strengths of LSTM and MLP models to leverage both the sequential dependencies of input features and the parallel processing capabilities of MLP. The LSTM model captures the temporal dynamics and dependencies within the loan data by employing Recurrent Neural Network (RNN). . The Hybrid deep LSTM-MLP acquired the maximum rate of accuracy of 97.78%, a rate of specificity of 93.92% and FNR of 0.026.