Kampo medicine is the Japanese adaptation of traditional medicine. In Kampo medicine, "medical interview" plays an important role. "Medical interview" in Japanese traditional medicine includes not only chief complaint but also a questionnaire that asked about the patient's lifestyle and subjective symptoms. The diagnosis by Kampo is called "Sho" and determined by completely different view from Western medicine. Specialists gather all available information and decide "Sho." And this is the reason why non-Kampo specialists without technical knowledge have difficulties to use traditional medicine. We analyzed "medical interview" data to establish an indicator for non-Kampo specialist without technical knowledge to perform suitable traditional medicine. We predicted "Sho" by using random forests algorithm which is powerful algorithm for classification. First, we use all the 2830 first-visit patients' data. The discriminant ratio of training data was perfect but that of test data is only 67.0%. Second, to achieve high prediction power for practical use, we did data cleaning, and discriminant ratio of test data was 72.4%. Third, we added body mass index (BMI) data to "medical interview" data and discriminant ratio of test data is 91.2%. Originally, deficiency and excess category means that patient is strongly built or poorly built. We notice that the most important variable for classification is BMI.