Rough set enabled classifiers for uncertain data

Abstract

Uncertainty in real world data makes the available knowledge imperfect and reduces the efficiency of any data mining or decision making task Particularly in decision making systems that involve multiple categories or decision classes uncertainty arises when the input pattern is ambiguous and when the classes are overlapping or ill defined Existing classifiers that partake extensively in many real world applications lack intelligence in handling these uncertain small sample sized and inconsistent data sets Hence there is a need for an intelligent classification system that perform well despite the occurrence of uncertainties This thesis work makes an attempt to build intelligent classifiers that address uncertainty issues Initially a study and analysis of uncertainty in some data sets was carried out using some data mining techniques Taking cues from this this research work goes on to build classifiers that handle rough and fuzzy uncertainties in the data sets. newline