A Heuristic Algorithm to Attribute Reduction with Simple Common-cost

Abstract

Cost-sensitive learning is an important issue in both data mining and machine learning. Most existing research works focus on decision systems where test-costs are additive. However, in some applications there is a common-cost within a group of tests. In this paper, we design an attribute reduction algorithm with a heuristic function and a parameter adjusting scheme to deal with this situation. The heuristic function has two parameters serving as the reward and the penalty exponent, respectively. The parameter adjusting scheme is based on the competition approach. Experimental results on four UCI (University of California-Irvine) datasets indicate that the algorithm obtains optimal reducts in most cases.