Abstract:It is valuable to recognize the correlation between the skin state index of human face and the type of Traditional Chinese Medicine (TCM) constitution with scientific and quantitative research methods, and from the process perspective of the accumulation of data and the further advance of knowledge discovery in database, the complex dynamic evolution law of TCM constitution and the appearance skin state index would be revealed. A classification model was proposed that combined the good inductive properties of decision tree for small sample data and the high classification accuracy of Bayes algorithm for large sample data. The decision tree and fuzzy Naive Bayes algorithm were fused to optimize adaptively the weight of index under the trend of the test data accumulation from less to more with a better classification accuracy and interpretation performance. The post-pruning way was used to avoid the over fitting of the conventional decision tree. The fuzzy membership function was introduced to the Naive Bayes by defining interval boundary to solve the randomness and fuzziness in testing and classification of skin state index. The results showed that the fusion weights of the proposed classification model could be adjusted dynamically and the classification accuracy would be increased with the increase of modeling data. The classification accuracy of the classification model with 151 data is 86.7%, higher than 83.3% of independent decision tree, 80% of Naive Bayes,and 76.7% of the matched group with 80 data. The dynamic classification model of facial skin and TCM constitution could effectively identify the complex relationship between the facial skin index and the internal type of TCM constitution by utilizing the data information involved in modeling. The classification model has good accuracy and interpretability, which is a beneficial exploration for the scientific and intelligent development of TCM theory based on data-driven.