An examination and comparison of conflicting data in granulized datasets: EWI vs. EFI

Knowledge discovery in Databases (KDD) frequently faces the need of making the granulized dataset consistent for the continuous database. Common techniques used in granulization are equal width interval (EWI) and equal frequency interval (EFI). However, they may produce different results in terms of the number of conflicting records in a granulized dataset. In Wu´s research, the data inconsistency has been found significant in granulized datasets by using EWI. Following this finding, our research conducts an experiment to examine and compare the performance of utilized EWI and EFI. Eighteen continuous datasets are examined. The TCRM model that is introduced by Wu, embedding a technique of database structured query language is utilized to efficiently derive results. Experimental results obtained indicate that (1) of 18 datasets that were examined, 7 were not recognized via EWI while 8 via EFI, implying that almost 40% of the granulized datasets contained conflicting data; (2) comparatively, we did not find a notable difference between EWI and EFI with respect to their granulization performance, and (3) no remarkable tendency of conflicting data production against dataset size, the number of attributes, and the number of classes was found.