Stock Price Pumping and Dumping Manipulation Detection Using Deep Learning Methods in Tehran Stock Exchange

One of the most important challenges in the stock market is the manipulation of securities. Manipulation of securities not only causes problems in the functionality of the capital market in the assets pricing, but also causes damages to the investors’ confidence. Therefore, the design and use of smart surveillance tools, especially tools based on artificial intelligence, to detect the market manipulation is very important. The purpose of this research is to provide a model based on deep learning methods to detect stock pump and dump manipulation. For this purpose, we generated labeled manipulation data using designing surveillance triggers at Securities and Exchange Organization and checking suspicious transactions. In the following, by using the adversarial generative network, misbalancing of data has been solved, and by using multi-layer perceptron neural network models, recurrent neural network, long-short-term memory neural network, and radial basis function network, price manipulation has been detected. The results show that long-term memory network and recurrent neural network had a good performance at price manipulation detection by more than 92% F2 score and have the ability to be implemented in regulatory processes and future research in other areas of stock manipulations.