A novel method for quantitative diagnosis based on decision tree in Traditional Chinese Medicine

Traditional chinese medicine (TCM) is one of the most important complementary and alternative medicines. In this paper, a novel computerized diagnostic method based on decision tree (DT) is proposed for promoting standardization and popularization of TCM diagnosis. In TCM, the symptoms are often high dimensional. Although DT induction algorithm has a feature selection scheme included in its learning performance, but this scheme is not optimal. The redundant and irrelevant symptoms may degrade the performance of the induced classifier. In this work, we utilize feature selection algorithm prior to the learning phase. The experiments show that the proposed method constructs much simpler tree and obtains relative reliable predictions. The rate of predictive accuracy in diagnosing apoplexy reaches 94.15%. The results suggest that the method proposed is feasible and effective and can be expected to be useful in the modernization of TCM.