Modeling traditional Chinese medicine doctor's diagnosis based on the multimodal dataset with subjectivity

This paper proposes a methodology for modeling a traditional Chinese medicine (TCM) doctor's diagnosis by classifying a person's health states into 13 Zhengs that are not entirely independent, while the diagnosis data as the decision data are subjective. A method for selecting the optimal modals and the attributes from the multimodal input data for the classification is proposed, and the corresponding index for the validation is defined. Moreover, the quality of the diagnosis from the TCM doctor is estimated by the defined index, and reliable samples for training the Zheng classifiers are selected based on the diagnosis data. The simulation and the prototype experiments show the effectiveness and efficiency of the proposed methodology.