Analysis of the characteristics affecting the trading risk of listed companies' stocks: A hybrid spatial artificial intelligence approach

This research identifies the determinants of trading risk (conditional variance) in the Iranian stock market over ۱۵ years (۲۰۰۸-۲۰۲۳) using a novel hybrid approach combining spatial econometrics and machine learning algorithms. The main objective is to evaluate the superiority of hybrid models over traditional methods and to identify the roles of macroeconomic, geopolitical, behavioral factors, and firm characteristics in shaping systematic risk. The sample includes ۱۷۲ companies listed on the Tehran Stock Exchange with ۳۰,۹۶۰ monthly observations and ۳۳ explanatory variables. The methodology was implemented in three stages: First, GARCH and EGARCH models were employed to extract conditional variance and confirm the leverage effect. Second, the Spatial Durbin Error Model (SDEM) was used to decompose direct, spatial spillover, and total effects of variables while controlling for cross-sectional dependence (Pesaran CD statistic = ۸۷.۴۵***) and spatial autocorrelation (Moran's I = ۰.۴۵۶۷***). Third, machine learning algorithms, including Linear Regression, SVM, Random Forest, XGBoost, LSTM, and Transformer, were applied independently and in combination with SDEM outputs. The results demonstrated a clear performance hierarchy: Linear Regression (R² = ۰.۴۱۲۳, RMSE = ۰.۰۹۸۷), SVM (R² = ۰.۵۹۸۷), Random Forest (R² = ۰.۶۷۸۹), XGBoost as the best standalone model (R² = ۰.۷۴۵۶, RMSE = ۰.۰۵۳۴), and Ensemble (R² = ۰.۷۵۲۳). Hybrid models showed significant superiority: SDEM + XGBoost (R² = ۰.۷۸۲۳, RMSE = ۰.۰۴۷۱; ۱۱.۸۰% error reduction compared to standalone XGBoost and ۵۲.۳% improvement over Linear Regression), and SDEM + Ensemble (R² = ۰.۷۸۶۷, RMSE = ۰.۰۴۶۷) achieved optimal performance. Time-series cross-validation (average test RMSE = ۰.۰۴۹۲) and the Diebold-Mariano test (DM = ۳.۴۵۶*** against XGBoost) confirmed statistical superiority. From a substantive perspective, the exchange rate with a total effect of ۰.۲۴۴۳*** and SHAP contribution of ۱۸.۳۴% was identified as the most important systematic risk factor, followed by sanction intensity (total effect = ۰.۱۲۷۴***, SHAP = ۱۴.۲۳%), Altman Z-score (SHAP = ۱۵.۶۷%), total stock index (total effect = -۰.۱۸۰۱***, SHAP = ۱۲.۸۹%), and investor sentiment (total effect = ۰.۱۰۰۱***, SHAP = ۱۱.۴۵%). The findings demonstrate that hybrid spatial econometrics and machine learning models improve prediction accuracy by ۱۲-۱۵% through extracting complementary information. Geopolitical and behavioral factors, in addition to traditional macroeconomic variables, are systematically important. Spatial spillovers constitute ۱۵-۲۵% of total effects, which are ignored in traditional models. This research shows that the frontier of financial risk modeling lies in the synergistic integration of economic theory and machine learning.

کلیدواژه ها:

Trading Risk ، Spatial Econometrics ، Machine Learning ، Iranian stock market