A rough set theory based predictive model for stock prices

Attempting to successfully and accurately predict the financial market has long attracted the interests and attention of economists, bankers, mathematicians and scientists alike. The financial markets form the bedrock of any economy. There are a large number of factors and parameters that influence the direction, volume, price and flow of traded stocks. This coupled with the markets' vulnerability to external and non-finance related factors and the resulting intrinsic volatility makes the development of a robust and accurate financial market prediction model an interesting research and engineering problem. In an attempt to solve this engineering problem, the authors of this paper present a rough set theory based predictive model for the financial markets. Rough set theory has, as its base, imperfect data analysis and approximation. The theory is used to extract a set of reducts and a set of trading rules based on trading data of the Johannesburg Stock Exchange (JSE) for the period 1 April 2006 to 1 April 2011. To increase the efficiency of the model four dicretization algorithms are used on the data set, namely, Equal Frequency Binning (EFB), Boolean Reasoning, Entropy and the Naïve Algorithm. The EFB algorithm gives the least number of rules and highest accuracy. Next, the reducts are extracted using the Genetic Algorithm and finally the set of dependency rules are generated from the set of reducts. A rough set confusion matrix is used to assess the accuracy of the model. The model gives a prediction accuracy of 80.4% using the Standard Voting classifier.