Recognition of TCM syndrome types of cardiovascular diseases based on support vector machine and artificial neural networks

Objective To discuss the recognition method of TCM syndrome types of cardiovascular diseases,and provide some methods and evidences for the standardization of TCM syndrome research.Methods The cases of cardiovascular diseases were selected by using unified scale of TCM inquiry.A database of TCM clinical information was established through assigning "yes" to 1 and assigning "no" to 0 according to the information by inquiry.The clinical information and relation among syndrome types were analysed and model was established based on support vector machine(SVM,including two calculation methods: radial basis function and polynomial kernel function) and artificial neural networks(ANN,including two structure networks: ACON and OCON).The accuracy of syndrome deduction of SVM was observed.Results For heart-qi deficiency,heart-yang deficiency,heart-yin deficiency,phlegm turbidity,qi stagnation and blood stasis,common syndrome types of cardiovascular diseases,the recognition accuracy of OCON structure network was the highest(over 60%),and reached 92.4% and 82.9% respectively for heart-qi deficiency and heart-yang deficiency.Conclusion SVM and ANN can provide some objective bases for the recognition of TCM syndromes of cardiovascular diseases,and OCON structure network has a higher recognition accuracy.