Evaluating industry performance using extracted RGR rules based on feature selection and rough sets classifier

In strategy of investment, an important thing for investors is to correctly predict firm's revenue growth rate (RGR), which is an effective evaluation indicator for them to see how big the potential power of future development is and measure how about: the growth of future development for a target firm that may be selected to investment portfolios. However, conventional methods of forecasting RGR have some shortcomings such as statistical methods based on strict assumptions of linearity and/or normality limit applications in real world. Additionally, due to rapid changing of information technology (IT) today, some techniques (i.e. rough sets and data mining tools) have become important research trends to both practitioners and academicians. With these reasons above, a new procedure, using the feature selection method and rough sets classifier, is proposed to extract decision rules and improve accuracy rate for classifying RGR. In empirical study, an actual RGR dataset collected from publicly traded company of stock markets is employed to illustrate the proposed procedure. The experimental results of RGR dataset analyses indicate that the proposed procedure surpasses the listing methods in terms of both higher accuracy and fewer attributes, and the output of proposed procedure is to generate a set of easily understandable decision rules that are readily applied in knowledge-based investment systems by investors. (C) 2008 Elsevier Ltd. All rights reserved.