Comparison of Algorithms for Computing Data Set Reducts

This paper compare four different algorithms for computing reducts or a reduct approximations based on Rough Set Theory. Reducts are used to create rule sets and then to classification of the test set. The data of patients suffering from heart disease is used to make algorithm's accuracy measure and create Receiver Operating Characteristic (ROC) curves. It has been concluded that out of these four methods of reduct calculation, classification through table rule set using covering algorithm yields best accuracy. Moreover statistical methods for algorithms comparison are presented. The ROSSETA software is used for reduct generation, rule set calculation and rough set classification. The R software environment is used for statistical computing and graphics generation.