Communications of the IIMA


In this paper, data from the Indian stock market is used to study the prediction accuracy of various classification techniques that can be used to identify market manipulation. The data contains information regarding price, volume and volatility of various stocks. Techniques like discriminant analysis, a composite model based on artificial neural network – genetic algorithm (ann-ga) and support vector machine (svm) have been used for classifying stocks into manipulated and non manipulated categories. It is observed that the support vector machine based technique gives the best classification accuracy among the three techniques.