| تعداد نشریات | 24 |
| تعداد شمارهها | 903 |
| تعداد مقالات | 8,081 |
| تعداد مشاهده مقاله | 15,203,876 |
| تعداد دریافت فایل اصل مقاله | 12,848,346 |
Stacked Intelligence: A Robust Ensemble Approach to Forecasting Big Tech Stock Prices in Turbulent Markets (2020–2025) | ||
| Iranian Journal of Economic Studies | ||
| دوره 15، شماره 1 - شماره پیاپی 29، 2026، صفحه 7-34 اصل مقاله (1.86 M) | ||
| نوع مقاله: Research Paper | ||
| شناسه دیجیتال (DOI): 10.22099/ijes.2026.55448.2097 | ||
| نویسندگان | ||
| Younes Nademi* 1؛ Majid Ebtia2؛ Ramin Khochiany1؛ Sayyed Mohammad Hoseini3 | ||
| 1Department of Economics, Ayatollah Boroujerdi University, Boroujerd, Iran. | ||
| 2Zagros Data Sciences Research Group, Ayatollah Boroujerdi University, Boroujerd, Iran. | ||
| 3Gahar Artificial Intelligence Research Group, Ayatollah Boroujerdi University, Boroujerd, Iran | ||
| چکیده | ||
| Accurate forecasts of the mega-cap technology stocks—Apple, Amazon, Alphabet, Meta and Microsoft—are vital for risk management and asset allocation. This study proposes a stacking ensemble that fuses Support Vector Regression (SVR), Random Forests (RF) and Extreme Gradient Boosting (XGBoost) as base learners, with a parsimonious linear meta-learner. We use daily OHLC data from 2 January 2020 to 11 March 2025, a span capturing the volatility of the COVID-19 shock and its aftermath. After differencing to ensure stationarity, the target variable becomes the daily change in closing price (Δ"Close" ). Models are trained on an expanding 80% window and tested on the final 20% of observations. Performance is assessed on strictly out-of-sample predictions using RMSE, MAE, MSE, and R^2. Across all five firms, the ensemble achieves the highest explanatory power (R^2≈0.80"-" 0.83) for predicting daily price changes and lowers RMSE by 8–15% relative to the best individual model. Friedman tests show these improvements are significant at the 1% level for Microsoft, Meta and Alphabet, and at 5% for Amazon; Apple shows no significant difference. The results indicate that combining heterogeneous learners curbs overfitting and exploits complementary nonlinear and temporal signals, producing stable forecasts during extreme market stress. The framework provides investors and policymakers with a validated AI tool for improving risk-return profiles in tech-heavy portfolios and offers methodological guidance for future financial-forecasting research. | ||
| کلیدواژهها | ||
| Stock price forecasting؛ Artificial intelligence؛ Stacking ensemble؛ Big Tech equities؛ Machine learning | ||
|
آمار تعداد مشاهده مقاله: 85 تعداد دریافت فایل اصل مقاله: 95 |
||