In order to reduce the instances influence, PSO-based attribute selection method is proposed in this paper. This method mainly use PSO algorithm to solve the optimal entropy of instances and obtain the corresponding attribute threshold value. According to the threshold, the attribute priority is determined. At last, selecting attribute according to priority. In experiment, specify the concept attribute priority in ontolgoy and verify the algorithm performance. Experimental results show that this algorithm reduces the dependency to instances and improve the accuracy. Otherwise, the computation quantity is reduced.