In order to calculate a minimal reduct based on a heuristic reduction algorithm, the character of attribute repulsion related to the minimal reducts is analyzed and an attribute repulsion matrix is proposed. On this basis, by combining some classical addition and deletion methods, two improved heuristic reduction algorithms based on the attribute repulsion matrix are presented. The experimental results on some UCI(UC Irvine) machine learning data sets show that the proposed attribute repulsion matrix can completely improve the quality of reduct and is helpful for a heuristic algorithm to calculate the minimal reduct.