Investor Sentiment, Audit Quality, and Stock Returns: An AI-Driven Approach

The objective of the present research is to investigate the impact of investor sentiment on stock returns, emphasizing the moderating role of audit quality, using artificial intelligence methods within the period of ۲۰۱۲ to ۲۰۲۴. In this research, financial data of manufacturing companies listed on the Tehran Stock Exchange were collected and analyzed using six machine learning models, including Artificial Neural Networks (ANN), Multilayer Perceptrons (MLP), Decision Tree, Random Forest, Cerebellar Model Articulation Controller (CMAC) neural networks, and Gradient Boosting methods. The variables under investigation included investor sentiment (SENT), audit quality (AQ), the interaction variable (SENT × AQ), and Fama-French market risk factors. Multiple performance metrics and cross-validation methods were used to evaluate the models. The results demonstrate that investor sentiment is the most significant predictor of stock returns across all AI models, with importance levels ranging from ۲۸.۳% to ۳۳.۸%, particularly achieving ۳۳.۸% in Gradient Boosting and ۳۱.۵% in Random Forest models. Audit quality emerged as the second most critical variable with importance levels of ۲۳.۵% to ۲۶.۴% across different models. The interaction variable (SENT × AQ) showed substantial moderating effects with importance ranging from ۱۸.۶% to ۲۵.۴%, reaching ۲۴.۷% in Decision Tree models, providing strong statistical evidence for the moderating role of audit quality. Gradient Boosting methods achieved superior performance with R² = ۰.۹۱۱ and prediction accuracy of ۹۳.۸%, followed by Random Forest with R² = ۰.۸۹۲. Cross-validation results (۵-fold, ۱۰-fold, and time-series) confirmed model stability and generalizability.

کلیدواژه ها:

Artificial intelligence ، Audit Quality ، Investor Sentiment ، Fama French model ، Machine Learning ، Stock Returns