| ObjectiveIn recent years, the spread of AIDS has entered a new period of rapid growth. Chinese Medicine on the prevention and control of AIDS work reflects certain advantages. An accurate diagnosis of the type of syndrome in TCM (Traditional Chinese Medicine) syndrome types is the basis of the use of traditional Chinese medicine, but now there is not a rapid and effective diagnostic technique that could be used to diagnosis Chinese Medicine Syndrome Types. It is a challenge for Medical statistics to select a practical statistical method to build a diagnosis model which cans diagnostic the TCM syndrome type of HIV/AIDS faster and more accurate. The common traditional models of diagnosing a disease are regression models. The data of this kind of model is very strict, for example require the data material normal, linear and other conditions, but the data from four TCM diagnostic methods of AIDS patients and the data from clinical index is not satisfied, the information from four TCM diagnostic methods were classification variable, and there may be a non-linear relationship between them, so we have to consider another model which is more suitable for HIV/AIDS patients data. The data mining technology of the neural network technology and decision tree technology just to make up for the lack of regression analysis, as the main method of this study, This study want to through the information from four TCM diagnostic methods and clinical index for HIV/AIDS patients with major empirical modeling and main deficiency syndrome modeling, and traditional Chinese medicine diagnosis and treatment for HIV/AIDS patients to implement the technology is widely primary health institutions to provide the theory basis.MethodIn recent years, data mining is a new technology, both at home and abroad, its application of health care or a new field method, this research data from The10th Five Years Key Programs for Science and Technology Development of China. The study a multi-stage random sampling methods extraction research object, and1277cases of research object did a questionnaire survey, at last, we got253patients that they finished laboratory tests and questionnaire survey. At first, we describe the empirical and deficiency syndrome of253cases of patients, at last, we got173cases were diagnosed as main empirical object of study and the142cases were diagnosed as main deficiency syndrome object of study used for modeling analysis, according to75%for training set and25%for the testing set to divide the set of data. Which train set for neural network and decision tree model fitting, and the test set used to evaluate final neural network model and decision tree model. Analyses the data processed with SPSS16.0and SPSS Clementine.ResultIn this study, we descriptive the data, and selecting the main empirical and main deficiency syndrome of HIV/AIDS patients and create the model. Normal, internal accumulation of damp-hea and evil "skin accounting for68.4%of the TCM empirical, the rest of the type accounting for32.6%.In this study, we found the model by normal and internal accumulation of damp-hea and evil "skin accounting. Syndrome of the spleen and lung deficiency and spleen qi deficiency accounting for56.1%, the rest of the type accounting for43.9%, we found the model by syndrome of the spleen and lung deficiency and spleen qi deficiency.Use the feature selection node of Clementine select the main laboratory information and the information from four TCM diagnostic methods, we got65index. There were9invariables enter the empirical model, including Itchy skin, white tongue, yellow tongue, aversion to cold, thick tongue, thin tongue, greasy tongue, flustered and palpitation, the content of CD4, the correct rate of Neural network model was84.86%in the training set and73.81%in the testing set, the correct rate of decision tree model was81.203%in the training set and73.81%in the testing set. There were10invariables enter the model, including fever, cough and expectoration, vomiting, taste, chest tightness and chest pain, itchy skin, pink tongue, weak white tongue, the content of CD4, the correct rate of Neural network model was87.25%in the training set and80.00%in the testing set, the correct rate of decision tree model was82.35%in the training set and75.00%in the testing set. After evaluation, the main empirical and main deficiency syndrome data mining models were good prediction effect.DiscussionThis study selects the main laboratory information and the information from four TCM diagnostic methods through the feature selection node of Clementine and we got65index. Finally there were9variables into the main index of the empirical data mining model and10variables into the main deficiency syndrome data mining model.The diagnosis model made by neural network technology and decision tree technology has made a good diagnosis effect, this provide the theory basis for traditional Chinese medicine widely used to diagnosis and treatment for HIV/AIDS patients. |
PDF Full Text Request