Decision Tree Analysis With Feature Selection And Its Application In Diagnosis Of Traditional Chinese Medicine

As a part of the medicine treasury in the world, the traditional Chinese medicine science has been remarkably contributing to the health of people in China and around the world throughout about two thousand years. However, it also has been badly hindered from being popularized and further developing due to the unquantifiable and obscure features of its diagnostics. Therefore, it is the issue to be studied in this article on how to standardize and objectify the syndrome differentiation of traditional Chinese medicine science, which is full of obscurity. With the application of Decision Tree analysis, which is combined with Feature Selection technique, this paper is trying to find out the correspondence between symptoms and syndrome by studying the data of a hepatocirrhosis case. The objective is to find out a rule in the syndrome differentiation, which will probably provide reference to medical workers and make it possible for the digitalization and objectification of the diagnosis in traditional Chinese medicine science.Feature Selection is an important technique of data pre-processing, which is aimed to recognize so as to eliminate the features, in all data attributes, which are redundant or irrelevant to the issue being studied. Due to the high cost of data collection in the case of Chinese medical treatment, the data classification in Chinese medicine science is a typical small sample problem. Furthermore, for a Chinese medicine case, there are relatively more dimensions in data, thus there are more redundant and irrelevant parts and therefore, Feature Selection of data becomes really important if a more accurate result and rule of syndrome differentiation is to be worked out.Classification is the essence of syndrome differentiation in Chinese medicine science. There are a lot of methods for classification, among which Decision Tree algorithm, as an example-based inductive learning algorithm, stands out to become an effective tool for the construction of syndrome differentiation model and rule extraction, owing to its advantages of being able to extract rules more clearly, better process non-numerical data, show important decision attributes and classify more accurately, etc.Grounded on previous researches and by focusing on Feature Selection technique and Decision Tree classification model, this paper puts forward a decision tree analytical method combined with feature selection, which is applied in syndrome differentiation in the case of Chinese medical treatment for hepatocirrhosis, thus providing a new approach to the objectification of Chinese medical diagnosis. The work involved in this paper includes:First of all, it studies Feature Selection algorithms and analyzes the main purpose and process of Feature Selection, as well as the correlation-based feature attribute search method and features combination evaluation strategy. It brings forward A-FCBF, a Feature Selection algorithm with adapting capacity, thus simplifying the processing of data reduction and its accuracy as well.Secondly, it summarizes the advantages and disadvantages of commonly used Decision Tree algorithms and proposes E-ID3 Decision Tree algorithm, which is able to improve data reduction, decision feature selection and pre-pruning strategy. Compared to the original algorithm, the new one enhances the efficiency of processing, precision of predicting and understandability of the rules. In conclusion, the analytical algorithm discussed in this paper will be applied to the syndrome differentiation in Chinese medical treatment for hepatocirrhosis. By summarizing clear rules in syndrome differentiation, it can provide a reference to the objective diagnosis in Chinese medicine science, as well as making it feasible for the intellectualization of syndrome differentiation in Chinese medical treatment.