Stock Market Manipulation Detection using Artificial Intelligence: A Concise Review

In the modern era, one of the fundamental trading components of any country is its stock market. Generally, it is the benchmark used to gauge the health of the country's economy. Due to its utmost importance, governments all over the world have set up a set of rigorous regulations to ensure fairness in the trade market. However, there will still be a small group of rogue traders that try to manipulate the stock price through various illegal means. Several popular manipulation schemes such as pump and dump, layering, and quote stuffing are regularly employed to inflate and deflate the stock price artificially without any strong basis. These illegal activities will undermine the integrity of the stock market, which will cause big losses to honest traders. At the moment, a simple rule-based system is used by most of the stock exchange authorities to identify and recognize manipulative cases. More often than not, some of the manipulation cases might be left undetected, especially in the digital era, where the trading volume has increased significantly. Therefore, an automated tool is needed to supplement human operators in identifying and recognizing such manipulation cases. Nevertheless, the complexity of the manipulation strategies has also increased with the help of various automation tools. Hence, several works have explored the application of artificial intelligence technology to detect manipulation cases in the stock market trading automatically. This paper will first provide a brief introduction to stock market manipulation, followed by some literary works on artificial intelligence methods that have been applied to detect manipulation activities in the stock market. Most of the existing works rely on the standard machine learning approach that includes artificial neural networks, support vector machine, random forest, and decision tree classifiers. At the moment, there is no standard public database that can be used to benchmark the effectiveness of the machine learning techniques in detecting the manipulation cases. Each researcher focuses on stock exchange data from their country of origin, either through daily or tick trading data. At the end of this paper, some suggestions are shared for possible future research directions.