|
Stock Manipulation Detection using artificial intelligence models in the Iranian capital market
|
Hamed Hamedinia *   , Reza Raei , Ali Amirshahi |
| Department of Finance, Faculty of Management, University of Tehran , hamedinia.hamed@ut.ac.ir |
|
|
Abstract: (8 Views) |
Stock manipulation poses a major challenge in capital markets, threatening efficiency and investor confidence. This study aims to identify price manipulation in stocks listed on the Iranian capital market using artificial intelligence models. The dataset includes daily information for 69 stocks across 19 industries from 2016 to 2020. Due to the lack of transparent identification of manipulated stocks in Iran, synthetic data were generated by injecting random patterns into non-suspicious stocks, supported by expert judgment. Key variables examined include daily returns, trading volume, and the average buy and sell activities of individual and institutional investors. The research adopts an applied, empirical, and descriptive–analytical approach, utilizing AI models such as logistic regression, decision trees, and multilayer neural networks.
Findings show that manipulation periods are associated with significant increases in returns, trading volume, and investor activity. Among the models, the decision tree achieved the best performance, with 68% accuracy based on the F2-measure. It also identified the most influential indicators of manipulation, notably increased trading volume on and before the manipulation day, along with abnormal returns. Finally, the study provides practical, step-by-step guidelines to help regulators detect manipulated stocks more effectively. |
|
| Keywords: Stock Manipulation, Deep Learning, Artificial Intelligence (AI) |
|
|
Full-Text [PDF 1066 kb]
(12 Downloads)
|
|
Type of Study: Research |
Subject:
Special
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
