A New Efficient Hybrid Boruta-CNN-GA Technique for Predicting Financial Distress

Financial distress prediction (FDP) has become an important area of study since it can warn the organization, investors, and creditors early. This study presents a hybrid methodology for predicting financial distress using a convolutional neural network (CNN) and a Genetic Algorithm (GA) with Boruta automated feature selection. Our CNN-GA is based on GA-based tuning of the essential CNN hyperparameters-Network depth, Dense layer activation function, Network width, and Network optimizer. We then investigate the influence of Boruta feature selection on our CNN-GA algorithm. The prediction performance is assessed using samples of firms from with samples of companies from GCC countries. The experiments in this paper use resampling with k-fold evaluation metrics, ensuring a comprehensive evaluation of our approach. The simulation results show that adopting the Boruta method has consistently boosted the accuracy of our algorithm.