Multi-label Learning Method Based On Integrated Learning And Rule Extraction In The Dialectical Study Of Hypertension Syndrome

Hypertension is one of the important causes of cardiovascular diseases.In China,the prevalence of hypertension is showing an upward trend.Syndrome elements differentiation will receive great attention in the field of traditional Chinese medicine(TCM).TCM treats hypertension after differential diagnosis.With the accumulation of clinical hypertension data,it will be an interesting direction.Differential diagnosis is a process that the analysis of symptoms inferred syndrome elements then syndrome elements are synthesized into syndrome.TCM practitioners are mainly dependent on the lessons of previous experience.With the dramatic increase in TCM clinical data,this approach is not suitable for discovering new knowledge of TCM diagnosis.In this paper,we study syndrome elements differentiation by multi-label ensemble learning,unbalanced problem handling and rule extraction.We have done out works as follow:Firstly,we applied multi-label learning to syndrome elements differentiation for its sample consist of multiple symptoms and syndrome elements.We have also found that syndrome elements differentiation has the unbalanced problem.Secondly,we constructed a classifier with high performance for TCM syndrome elements differentiation by multi-label learning,ensemble learning and unbalanced problem process.In this paper,we used multi-label ensemble learning to find a suitable classifier model for syndrome elements differentiation.To solve the problem of unbalanced problem,we proposed a under-sampling method based on multi-label relevance.At last,we constructed a model with the balance between accuracy and interpretability for syndrome elements differentiation.The path from the root to the leaf of decision tree can be considered as a rule in the random forest.We trained random forest on the dataset of symptoms and syndrome elements and we used hill climbing to extract rule set of random forest for the guide of hypertension syndrome elements differential.In this paper,we did our job from high performance and high interpretability,and we have proposed two models for hypertension syndrome elements differentiation.These workscan bring the auxiliary decision and find rules for the diagnosis of hypertension in TCM,which benefited the Chinese medicine information,objective and modernization...