ANN-GA based model for stock market surveillance

• Published in: Journal of financial crime 2013, Vol.20 (1), p.52-66
• Main Authors: Punniyamoorthy, Murugesan, Joy Thoppan, Jose
• Format: Article
• Language: English
• Subjects: Capital markets; Classification; Data mining; Discriminant analysis; Efficient markets; Investments; Investors; Neural networks; R&D; Research & development; Scandals; Securities markets; Stock exchanges; Stock prices; Studies; Surveillance; Volatility
• ISSN: 1359-0790; 1758-7239
• DOI: 10.1108/13590791311287355

Purpose - This paper attempts to develop a hybrid model using advanced data mining techniques for the detection of Stock Price Manipulation. The hybrid model detailed in this article elucidates the application of a Genetic Algorithm based Artificial Neural Network to classify stocks witnessing activities that are suggestive of potential manipulation.Design methodology approach - Price, volume and volatility are used as the variables for this model to capture the characteristics of stocks. An empirical analysis of this model is carried out to evaluate its ability to predict stock price manipulation in one of the largest emerging markets - India, which has a large number of securities and significant trading volumes. Further, the article compares the performance of this hybrid model with a conventional standalone model based on Quadratic Discreminant Function (QDF).Findings - Based on the results obtained, the superiority of the hybrid model over the conventional model in its ability to predict manipulation in stock prices has been established.Research limitations implications - The classification by the proposed model is agnostic of the type of manipulation - action-based, information-based or trade-based.Practical implications - The market regulators can use these techniques to ensure that sufficient deterrents are in place to identify a manipulator in their market. This helps them carry out their primary function, namely, investor protection. These models will help effective monitoring for abnormal market activities and detect market manipulation.Social implications - Implementing this model at a regulator or SRO helps in strengthening the integrity and safety of the market. This strengthens investor confidence and hence participation, as the investors are made aware that the regulators implementing market manipulation detection techniques ensure that the markets they monitor are secure and protects investor interest.Originality value - This is the first time a hybrid model has been used to detect market manipulation.