Comparison between rough set theory and logistic regression for classifying firm’s performance

• Main Authors: Bahtiar Jamili Zaini,, Siti Mariyam Shamsuddin,, Saiful Hafizah Jaaman ,
• Format: Article
• Published: Penerbit ukm 2008
• Online Access: http://journalarticle.ukm.my/1862/; http://www.ukm.my/~ppsmfst/jqma/index.html

Superior prediction and classification in determining company’s performance are major concerns for practitioners and academic researchers due to the importance of providing useful information to shareholders and potential investors. The general practice is to analyze firm’s performance based on financial indicators measured on the financial statements published in the annual report, i.e. the balance sheet, income and cash flow statements. To do so, numerous financial indicators are used in order to classify the performance of firm accordingly involving many financial variables in order to determine firm’s performance. Some of these financial ratios may be irrelevant and may correlate with each other giving redundant information for classification. Hence, this study investigates and determines the financial ratios that notably affect firm’s performance so as to predict future performance. In this study, firm is considered as high performance firm if its share returns perform above returns provided by Kuala Lumpur Composite Index. Using various financial ratios as the independent variables, this study investigates and determines the financial indicators that significantly affect firm’s share performance so as to predict its future performance. Financial performance predictions normally involve large information to explore, using traditional statistical methods such as multiple discriminant analysis or binary logistic regression analysis. Parametric statistical models are based on normality assumption requirements for the interpretation of the tests of significance, and if the data does not satisfy this assumption, the results obtained may be biased. However, it is noted that most financial ratios are skewed and non-normally distributed suggesting that non-parametric test is a superior alternative. Non-parametric model does not assume the data to have any specific characteristics. In this study comparison between rough set and logistic regression methodology is employed to identify the most significant indicators in classifying firm’s performance. It was found that rough set model could accurately predict 86% of the testing dataset whereas logistic regression model shows 61.6% accuracy rate