In this paper, we regard the significance of attributes defined from the approximation quality with respect to the partition and condition attributes set as heuristic information, and introduce the heuristic information into genetic algorithm based on optimizing initial population. A new modify operator is used for introducing the heuristic information so as to maintain the ability of classification of the attributes set. The operator is an embodiment of local research method using heuristic information. So the algorithm converges quickly and has global optimizing ability. The algorithm is also analyzed theoretically and the correctness and effectiveness of the algorithm are shown in the experiments.