Classifying Firms’ Performance Using Data Mining Approaches

Abstract

Superior prediction and classification in determining company’s performance are major concern for practitioners and academic research in providing useful or important information to the shareholders and potential investors for investment decision. Generally, the normal practice to analysed firm’s performance are based on financial indicators reported in the company’s annual report including the balance sheet, income and cash flow statements. In this work, a few popular and important benchmarking machine learning techniques for the data mining including neural networks, support vector machine, rough set theory, discriminant analysis, logistic regression, decision table, sequential minimal optimization and decision tree have been tested as to classify firm’s performance. The data mining techniques produce high classification rate that is more than 92%. This work also has reduced total number of ratios to be evaluated due to long processing time and large processing resources. Finally, the CA/TA, S/TA, E/TA, GM, FC, PBT/TA, and EPS have been considered for of the final reduced financial ratios. The results show that the 7 reduced ratios are comparable as the common 24 ratios. And to the still produce high classification rate and able classify the firm’s performance.