An Intelligent Bankruptcy Prediction Model based on an Enhanced Sparrow Search Algorithm

Abstract

Bankruptcy detection becomes one of the major subjects in finance. Indeed, for apparent reasons, several actors like shareholders or managers show more attention to the possibility of a firm’s bankruptcy. Subsequently, various researches are being conducted on the matter of bankruptcy prediction. Recently numerous research works have explored the application of machine learning (ML) techniques to bankruptcy prediction by having financial ratios as predictors. This article devises an Enhanced Sparrow Search Optimization with Deep Learning Enabled Bankruptcy Prediction (ESSODL-BP) model. The proposed ESSODL-BP technique involves the forecasting of the bankruptcy of a financial firm. To accomplish this, the ESSODL-BP technique primarily follows the Z-score normalization approach. Followed by, the bidirectional long short-term memory (BLSTM) model is designed to predict the bankruptcy status of a financial firm. Then, the ESSO algorithm is utilized for optimally tuning the hyperparameters related to the BLSTM model and also boosts the prediction performance to a maximum extent. The performance validation of the ESSODL-BP technique is tested using a benchmark dataset. The experimental outcomes reported better performance of the ESSODL-BP technique over other approaches.