| تعداد نشریات | 24 |
| تعداد شمارهها | 916 |
| تعداد مقالات | 8,202 |
| تعداد مشاهده مقاله | 15,918,410 |
| تعداد دریافت فایل اصل مقاله | 13,469,035 |
Evaluating The Effectiveness of Advanced Machine Learning Algorithms in Reducing Stock Trading Risk on The Iranian Stock Exchange | ||
| Iranian Journal of Economic Studies | ||
| دوره 14، شماره 2 - شماره پیاپی 28، 2025، صفحه 443-500 اصل مقاله (656.38 K) | ||
| نوع مقاله: Research Paper | ||
| شناسه دیجیتال (DOI): 10.22099/ijes.2025.54990.2083 | ||
| نویسندگان | ||
| Javad Zolfaghary Tabesh1؛ Babak Jamshidinavid* 1؛ Mehrdad Ghanbary1؛ Afshin Baghfalaki2 | ||
| 1Department of Accounting, Ker.C., Islamic Azad University, Kermanshah, Iran. | ||
| 2Department of Economics, Ker.C. Islamic Azad University, Kermanshah, Iran. | ||
| چکیده | ||
| The primary objective of the present study was to thoroughly assess the efficiency of advanced machine learning algorithms in managing and mitigating stock trading risk in the Iranian capital market. Utilizing financial, economic, textual, and behavioral data from 172 companies over a 15-year period (2008–2023), this research estimated trading risk using advanced time-series models and modeled it against a total of 33 independent variables across five main dimensions. The methodology involved comparing the performance of six algorithmic families, ranging from simple regression to deep neural networks, in two scenarios: continuous prediction and risk level classification. The analysis of the results revealed that linear models performed very poorly due to the highly volatile and non-linear structure of the Iranian market. However, in the realm of non-linear methods, deep neural networks demonstrated the best capability for accurate and continuous risk prediction by recording the lowest error rates; this success is attributed to the ability of these models to identify long-term temporal patterns and the complex interaction of variables. In the section on classifying stocks into risk levels, algorithms based on gradient boosting decision trees achieved the highest efficiency with an accuracy exceeding 83% and were suggested as ideal tools for early warning systems. The analysis indicated that the primary drivers of risk volatility are macroeconomic variables (such as exchange rates and inflation), the effects of which are amplified by investor behavioral factors. Ultimately, this research confirms the significant and decisive superiority of advanced machine learning models over traditional approaches in analyzing risk within the Tehran Stock Exchange. | ||
| کلیدواژهها | ||
| Stock Trading Risk؛ Machine Learning؛ Deep Neural Network؛ Gradient Boosting | ||
|
آمار تعداد مشاهده مقاله: 226 تعداد دریافت فایل اصل مقاله: 481 |
||