Variable Length Reduct Vs. Minimum Length Reduct-A Comparative study

Abstract

Generation of huge volume of data from various projects and business houses throws new challenges to data mining research community. Analysis of such data is utmost important but development of new mechanisms to handling the data lag behind with this massive growth, resulting a tremendous volume of data retained without being studied. In many applications, often it is difficult to know exactly which features are relevant for a particular task. The irrelevant or redundant attributes should be removed which efficiently reduces dimensionality of the system and thereby complexity of the systems. Rough set theory (RST), a new mathematical tool is applied for generating reduced attribute set called reducts, which is not unique. In the paper, two different attribute reduction techniques generating variable length reducts and minimum length reducts have been discussed and compared considering bench mark datasets.