Elderly dementia would be considered as one of serious social issues in near future inJapan. A nonlinear regression method by support vector machine (SVM） was appliedto search factors related to patient rates of 47 prefectures among 34 kinds of lifestylehabit factors. Fourteen kinds of related factors were obtained; depression, alcohol,hyperlipidemia, hobby, fruits, stress disease, high blood pressure, soy product, cereal,fresh fish, cooking oil, exercise, fresh vegetable, and diabetes. Depression is the mostimportant factors to patient rates, and the relative significance of the related factors tothe patient rates of elderly dementia is discussed on the basis of their sensitivities. Theinformation obtained could be used for serving as a reference to factors which shouldbe verified in cohort or case-control studies for clarifying the causes of elderly dementiain Japan.