Consistency based rules mining on sparse and diverse TCM sub-health diagnosis data

This paper proposes a method of consistency based rules mining on sparse and diverse data set derived from the sub-health diagnosis of TCM doctors, so as to realize the automatic inference of individuals´ sub-health state and their corresponding TCM syndrome. Because of the data´s bias given by doctors, a consistency detection algorithm to find out the feature sets that can fit the doctors´ diagnosis is presented, and the rule mining algorithm is instructed by it to forecast the sub-health state. Derivation accuracies before and after using the consistency detection algorithm are given by our experiments. The performance of the consistency detection algorithm is evaluated, and the limitation is analyzed.