Apply Feature Selection to the Integration of TCM and Western Medicine

Differentiation of syndromes of traditional Chinese medicine (TCM) mainly depends on the information obtained from four diagnosis methods. Now many physicochemical parameters are available in clinic. There exists great correlation between TCM syndromes and physicochemical parameters. The objective of the paper is to analyze the correlation between TCM syndromes and physicochemical parameters quantitatively and find the most informative physicochemical parameter combination which is useful in differentiation of syndromes. A novel definition of correlation degree based on Renyi's entropy is proposed which can measure the correlation between variables efficiently. Feature selection based on the correlation degree is used to find the most informative physicochemical parameters for assisting differentiation of syndromes.