This paper presents a study of the style of SonCi using many machine learning algorithms whose parameters are optimized by the results of experiments. At the same time, the reverse analyses is performed to get which words have the most effects on the decision making. This method can be used in analyzing the writing style of some author.