Fast attribute selection based on the rough set boundary region

Abstract

The problem of clustering exists in numerous fields such as bioinformatics, data mining, and the recognition of patterns. The function of techniques is to suitably select the best attribute from numerous contending attribute(s). RST-based approaches for definite data has gained significant attention, but cannot select clustering attributes for optimum performance. In this paper, the focus is on the processes that exhibit a similar degree of results to an identical attribute value. First, the MIA algorithm was identified as the supplement to the MSA algorithm, which experiences set approximation. Second, the proposition that MIA accomplishes lesser computational complexity through the indiscernibility relation measurement was highlighted. This observation is ascribed to the relationship between various attributes, which is markedly similar to those induced by others. Based on the fact that the size of the attribute domain is relatively small, the selection of such an attribute under such circumstances is problematic. Failure to choose the most suitable clustering attribute is challenging and the set is defined rather than computing the relative mean where it can only be implemented with a distinctive category of the information system, as illustrated with an example. Lastly, a substitute method for selecting a clustering attribute-based RST using Mean Dependency degree attribute(s) (MMD) was proposed. This involved selecting the maximum value of a mean attribute(s) as a clustering attribute through a considerable targeting procedure for the rapid selection of an attribute to settle the instability in selecting clustering attributes. Thus, the comparative performance of the selected clustering attributes-based RST techniques MSA and MIA was conducted.