Indicator Selection based on Rough Set Theory

A method for indicator selection is proposed in this paper. The method, which adopts the General Methodology and Design Research approach, consists of four steps: Problem Identification, Requirement Gathering, Indicator Extraction, and Evaluation. Rough Set approach also has been applied in the Indicator Extraction phase. This phase consists of 5 steps: Data selection, Data Preprocessing, Discretization, Split Data, Reduction, and Classification.. A dataset of 427 records have been used for experimentation. The datasets which contains financial information from several companies consists of 30 dependant indicators and one independent indicator. The selection of indicators is based on rough set theory where sets of reducts are computed from a dataset. Based on the sets of reducts, indicators have been ranked and selected based on certain set of criteria. Indicators have been ranked through computation of frequencies in reduct sets. The major contribution of this work is the extraction method for identifying reduced indicators. Results obtained have shown competitive accuracies in classifying new cases, thus showing that the quality of knowledge is maintained through the use of a reduced set of indicators.