Deep Learning-Based Model for Financial Distress Prediction.

Area of Science:

• Finance
• Machine Learning
• Computational Intelligence

Background:

• Financial distress prediction and credit scoring are critical in finance.
• Existing research primarily uses statistical and machine learning models.
• There is a need for improved accuracy in financial distress prediction models.

Purpose of the Study:

• To develop a novel financial distress prediction model using an adaptive whale optimization algorithm with deep learning (AWOA-DL).
• To enhance the predictive accuracy of deep neural network models through optimized hyperparameter tuning.

Main Methods:

• The study proposes the AWOA-DL technique, integrating a deep neural network (DNN) with an adaptive whale optimization algorithm (AWOA).
• A multilayer perceptron (MLP) forms the core DNN, with AWOA used for hyperparameter tuning.
• The model involves three stages: data preprocessing, AWOA-based hyperparameter tuning, and prediction.

Main Results:

• The AWOA-DL model demonstrated superior performance across four datasets.
• The proposed method achieved an average accuracy of 95.8%.
• Compared models achieved average accuracies of 93.8%, 89.6%, 84.5%, and 78.2%.

Conclusions:

• The AWOA-DL technique significantly improves financial distress prediction accuracy.
• This hybrid approach offers a robust solution for assessing corporate financial health.
• The findings highlight the potential of metaheuristic optimization combined with deep learning in financial analytics.