Investigating the impact of deep learning-based financial forecasting models on the accuracy of corporate profitability analysis

Abstract

Deep learning-based financial forecasting models are increasingly used to analyze corporate profitability. This study investigates the impact of such models on profitability prediction accuracy compared to traditional methods. Financial data from 60 Tehran Stock Exchange-listed companies from 2015 to 2024 were collected and analyzed using deep learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Gradient Boosting models. The findings indicate that deep learning models outperform traditional regression and time series models in profitability prediction accuracy. Additionally, integrating deep learning models with dimensionality reduction techniques enhances performance. However, computational complexity, data volume requirements, and high processing costs remain major challenges in implementing these models in financial environments. This study provides recommendations for optimizing the use of deep learning in profitability forecasting.