نمذجة محددات القيمة السوقية للأسهم باستخدام خوارزمية الغابة العشوائية دراسة تطبيقية على الشركات المدرجة في سوق دمشق للأوراق المالية

Abstract

This study investigates firm specific and market activity determinants of bank equity market value in the Damascus Securities Exchange using monthly data from 2015 to 2023. The design combines five operational inputs that are observable at high frequency earnings per share, trading volume, turnover, stock return, and trading days. A random forest with randomized hyperparameter search and time-based cross validation delivers strong out of sample performance R2 equals 0.8326 and RMSE equals 323.25. Variable importance ranks earnings per share first, followed by trading volume and turnover, with smaller yet nontrivial roles for stock return and trading days. A depth limited surrogate tree provides actionable thresholds linking profitability and trading activity to distinct market value clusters. Distributional diagnostics indicate heavy tails and intermittent trading that favor nonlinear, robust learning over linear benchmarks. The findings translate into a practical monitoring scheme that prioritizes sustaining profitability and activating liquidity through market making and disclosure cadence. The contribution is twofold evidence on bank equity valuation under severe liquidity frictions in Syria and a transparent, reproducible machine learning workflow that turns black box models into decision rules usable by issuers and regulators.