Abstract
Recently, there have been emerging research activities in classifying the Sasang constitution objectively by machine-based methods. The concordance rate of the classification by the pulse wave features was, however, only around 50% up to now. In this paper, we introduce a novel classification algorithm that can promote the accuracy substantially at the expense of the non-classifiable subgroup as a byproduct. For instance, with the pulse wave features alone, we show that female/male subject group in their 20s can be classified into the Sasang constitution group with the concordance rate of 68.4%/65.5% for a subgroup of 57/29(31%/15%) subjects out of 184/195, by leaving the other subjects as the non-classifiable group. Next, we show that the pulse diagnosis has been used only as a supportive tool in determining one's constitution, and consequently the accuracy of the concordance ratio by the pulse wave features alone cannot exceed a finite value, which we estimate to be about 60%.