Abstract: In this paper, discriminatory classification of the places, where earthquakes occur easily, is studied using Simming method, its improvement, and Bayes method. The results show that these three methods are more effective than the Cora-3 method. The features of preparatory process are selected by means of the characteristic of relation, and we found that the relations are better when the number of features is 20. The method of choosing the thresholds is improved using the iterative optimization method. In addition, the samples are classified by means of 2-dimension order clustering method and its results are compared with that of other methods.