Comparison of the Efficiency of Traditional Multivariate Regression and Modern AI-Based Optimization Algorithms in Predicting the Probability of Negative Stock Returns

Timely identification and prediction of negative stock returns represent a critical challenge in financial analysis and risk management, significantly influencing the effectiveness of investment decision-making. This study evaluates the traditional multivariate regression approach alongside the performance of advanced artificial intelligence (AI) optimization algorithms, including Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Ant Colony Optimization (ACO), in modeling the probability of negative stock returns. A comprehensive financial dataset was compiled from companies listed on the Tehran Stock Exchange. After conducting preprocessing and feature selection procedures, the models were evaluated based on performance indicators such as accuracy, recall, F1-score, and generalizability. Focusing on Iran's capital market, the research analyzed data from 101 publicly listed companies across two time periods (2010–2015 and 2023–2024). The objective was to assess the accuracy, stability, and generalizability of AI-based models in comparison to the classical regression model. The findings indicate that AI algorithms demonstrate superior performance over multivariate regression due to their ability to model complex, nonlinear relationships between financial variables. These algorithms showed higher predictive accuracy in detecting negative returns. Moreover, their capacity to manage large datasets and reduce statistical noise contributed to enhanced model stability when dealing with irregular and outlier data. These results suggest that employing AI-based optimization methods can serve as an effective tool for market risk analysis and improving decision-making processes under conditions of uncertainty. The present study offers a novel perspective on integrating traditional and intelligent techniques to enhance predictive accuracy in financial markets and may serve as a practical reference for financial analysts and policymakers.