Designing an Intelligent Leadership Model for Managers of Education Departments in Tehran Province: A Qualitative Grounded Theory Study

Objective: The study aims to develop an efficient predictive model for Iran’s capital market stress index by integrating machine learning algorithms with dynamic econometric modeling. Methodology: Daily time-series data of 20 industrial indices from the Tehran Stock Exchange over a ten-year period (2014–2024) were analyzed. Log returns were calculated, and normality, stationarity, and ARCH effects were tested. Systemic risk was measured using the DCC–MGARCH framework combined with the ∆CoVaR criterion. Three supervised machine learning models—Random Forest, Support Vector Regression, and Artificial Neural Networks—were compared to determine feature importance and extract index weights. The final stress index was constructed using dynamic conditional correlations and normalized risk weights. Findings: Most indices exhibited non-normality but were stationary and heteroskedastic. The Engle–Sheppard test confirmed dynamic conditional correlations. Based on ∆CoVaR, automotive, real estate, paper products, and basic metals demonstrated the highest systemic risk. Random Forest achieved the lowest MAE and RMSE, outperforming other models. The constructed stress index successfully identified high-risk periods and issued early warning signals prior to major market downturns. Granger causality tests revealed a unidirectional causal effect from the free-market exchange rate to the stress index, while gold coin returns showed no significant influence. Conclusion: The hybrid modeling framework—combining DCC–MGARCH with machine learning—provides a reliable, forward-looking tool for monitoring systemic risk and forecasting comprehensive market stress in Iran’s capital market.