Abstract: Decision tree is a highly effective method in data mining.However，the number of decision tree node will be exponential growth with the increament of attributes.Therefore，there are too many rules when the number of attributes is large.To avoiding this problem，the paper presents an optimization method based on genetic algorithm.Firstly，choose some groups of attribute with roulette wheel method based on information gain，construct decision trees with these attributes，and then，genetic algorithm is used to recombine them.So the rule set could be created with the result of the combination.Finally，the results of the experiments indicate that the method is feasible and effective.