Comparison of support vector machine, back propagation neural network and extreme learning machine for syndrome element differentiation

Abstract

Artificial intelligence is a discipline which focuses on the study of simulating and extending human intelligence, while machine learning (ML) is one of the most rapidly developing subfields of AI research. The research of ML in the field of traditional Chinese medicine (TCM), has a bright prospect, as well as a profound significance. To explore the feasibility of using ML methods to approximate the diagnosis of TCM, support vector machine (SVM) is introduced and investigated for syndrome element differentiation of TCM in this paper. Based on 670 medical records, SVM was used to approximate the mapping relations between inputs (set of clinical manifestations) and outputs (set of syndrome elements). The value orders of syndrome elements were adopted to evaluate the approximation results, while attribute partial order structure diagram was employed to discover and visualize the knowledge of the records. Back propagation neural network (BPNN) and extreme learning machine (ELM), as comparative methods, were also employed to deal with the medical data. The value order’s matching results between real and predicted results shows that, for SVM, the matched degree of each record is no less than 65%, while there are at least 88% records whose matched degree is more than 80%; and for BPNN and ELM, the highest proportion of records whose matched degree is more than 80% is only 45%. Using methods of ML to approximate the diagnosis of TCM should be feasible, and more relevant research can be conducted in the future.