Mutual information can measure arbitrary statistical dependencies. It has been applied to many kinds of fields widely. But when mutual information is used as the correlation measure, the features with more values are apt to be chosen. To solve this problem, a novel definition of correlation degree is proposed in this paper. It can avoid the shortcoming of selecting more value features when mutual information acted as the measure, and it can avoid the shortcoming of selecting less value features when correlation degree coefficients acted as the measure. In the method using the novel definition, the number of selected features is determined by the correct classification rate of Support Vector Machine. At last, the efficiency of the method is illustrated through analyzing the symptoms combination of seven essential elements of the syndrome corresponding to stroke.
Computing Methodologies ; Data Interpretation, Statistical ; Diagnosis, Differential ; Humans ; Medicine, Chinese Traditional ; methods ; standards ; Models, Statistical ; Stroke ; diagnosis