Induction as pre-processing

Abstract

In most data mining applications where induction is used as the primary tool for knowledge extraction, it is difficult to precisely identify a complete set of relevant attributes. The real world database from which knowledge is to be extracted usually contains a combination of relevant, noisy and irrelevant attributes. Therefore, pre-processing the database to select relevant attributes becomes a very important task in knowledge discovery and data mining. This paper starts with two existing induction systems, C4.5 and HCV, and uses one of them to select relevant attributes for the other. Experimental results on 12 standard data sets showtha t using HCV induction for C4.5 attribute selection is generally useful.