Investigations on the Suitability of Data Mining Techniques in Stock Market Turnover Prediction

Turnover prediction of a company in the ever fluctuating stock market has always proved to be a herculean task at hand. Data mining is a well-known research area of Computer Science that aims at extracting meaningful information from large databases. However, despite the existence of many algorithms for the purpose of predicting future trends, their efficiencies are questionable. The objective of this paper is to improve the accuracy of the prediction process by implementing various clustering, discretization and feature selection techniques on stock market data. The authorized Stock market dataset was taken fromwww.bsc.com and included the everyday stock values of various companies over the past 10 years. The algorithms were investigated using 'R' and `Weka' tool. To begin with, four clustering techniques-K-means clustering, Farthest first, Expectation Maximization and Canopy algorithm were used to divide the attributes into various clusters. After clustering, data discretization was performed using both supervised and unsupervised algorithms. A new data discretization algorithm was proposed in order to make the data mining investigations easier. Then, several feature selection algorithms like CFS Subset, Information Gain, Gain Ration, One R Attribute and Principal component Analysis were used to extract the important features. Finally, the Random Forest algorithm was utilized to predict the turnover. The above four steps were implemented to predict the turnover of a company on an everyday basis. An accuracy rate of 97% was achieved by the proposed methodology and the importance of the stock market attributes was established as well.