| Traditional Chinese Medicine(TCM)is an import part of traditional Chinese culture,and it play an indispensable role in the process of ancient Chinese people struggle with the disease.Even today,TCM play an important role of medical means as an important complementary of Western Medicine.By experience tremendous impact on formation of TCM theory,there are great uncertainties,limiting the further development of traditional Chinese medicine.The introduction of data mining technology,provides a new way for study of traditional Chinese medicine.Discover the objective laws from the large historical data to reveals the essence of the theory of traditional Chinese Medicine and to further promote the development of TCM research.This paper study Parkinson disease diagnosis by multi-label technology.Study multi-label algorithm and apply it in the practical application of Parkinson disease in TCM diagnosis to promote the application of traditional Chinese medicine in the diagnosis of Parkinson’s disease.The main work is as follows:1)Propose a density based weighted multi-label k-nearest neighbor algorithm,DWBML-KNN.The sample average similarity representation sample distribution density.Define an appropriate weighted factor,density of small sample receive larger weights,to reduce the adverse effects of sample distribution density imbalance problem.Finally,improve the performance of classification algorithm.2)For issues and challenges that exist in the TCM diagnosis of Parkinson’s disease,the proposed two-stage diagnostic prediction model,multi-label learning algorithm is applied to TCM diagnosis of Parkinson’s disease.Through joint action syndromes predicted and syndromes corrected two-stage,multi-label learning technology will be better applied in practice,and promote TCM in the diagnosis of Parkinson’s disease.3)Continue in-depth study of multi-label study will consider the relationship between the labels,propose a new idea that conditional probability measure label relations.And apply it to the ML-KNN algorithm to put forward a metric-based ML-KNN algorithm which measured by label conditional probability,LCPML-KNN.Experimental results show that the new idea can not only achieve good results in the public data sets,on historical data TCM diagnosis of Parkinson’s disease can still get a good result,to provide long-term technical support for multi-label learning better application in traditional Chinese medicine in the diagnosis of Parkinson’ s disease. |
