| Generally, Traditional Chinese Medicine(TCM) has been considered as an experience disciplines, therefore the research of standardization for TCM is a precondition that forwards to scientific and modernization. In recent years, with the development of Western medicine scale, some scholars have used the relevant knowledge and methods of Surveying and metrology in TCM research, which has been used extensively in the international community. Scale becomes one of the important means that accelerate the standardization of TCM. According to the information of the syndrome and vital signs in patients, the research of TCM Syndrome Scale Construction is an important part of TCM Clinical Diagnostic Scale Evaluation System, furthermore, it will hasten the standardization and scientific of TCM diagnosis.This thesis is based on the national key fundamental researches development plan project ("973 Project") "The Study on the Evaluation Criteria of the Diagnosis and Therapeutic Effects with the integration of Disease and Syndrome of Ischemic Stroke". By using the large sample of forward-looking information consultation survey and research data, feature selection method is introduced into the selection of scale items in this paper, which to raise the objectivity and accuracy of scale clinical diagnosis.The selection of scale items is not only the most crucial steps for the scale production, but also a researching hot topic in the current. Feature selection is defined that from the input characteristics of a set of evaluation criteria to select the optimal subset of the features. From the mathematical essence, the goal of the feature selection is consistent with the selection of scale items. Therefore, feature selection method can be applied to the selection of scale items. Through analyzing and recalling for the feature selection algorithm, the selection of scale items based on traditional and feature selection are compared. The main research in this paper is as follows:(1) After analyzing the background of the project, the targets of the study is determined and the major problems are put forward. Based on a brief overview on the scale principle, the application of the scale in TCM at present and the faced problems are introduced, especially for the selection of scale items, which is the most crucial part in the Establishing of scale. At the same time, the traditional methods of selection of scale items are studied and compared, and the method of factor analysis and stepwise discrimination analysis are described in detail. A method named improved factor analysis based on the stepwise discrimination analysis is proposed here, which is proved by the medical experiments simulation finally.(2) The feature selection methods are introduced. After summarizing the definition, the history, the current situation, the main course and the direction of the method of feature selection, two types of feature selection algorithm are introduced in details, which is called as Filter and Wrapper. A number of classic methods are also cited here. Therefore, a new feature selection algorithm combining the Filter and Wrapper two-phases is proposed.(3) In this paper, the method of feature selection is used in the selection of Apoplexy Syndrome Diagnosis Evaluation scale items. Emulation is presented by MATLAB 7.1 software. First of all, the data extraction is completed successfully, which is a guarantee for the following study. Secondly, through studying the similarity measure, four common distances are compared here, and we choose the best distance from the expertise of coincidence and classification accuracy rate. Thirdly, the application of Particle Swarm Optimization (PSO) algorithm is introduced to determine the weight of scale items; K-nearest neighbor method is chosen as the evaluation function. Finally, Experiments on 307 relatively high-dimensional datasets show that the algorithm has more comprehensive performance than the traditional methods with respects to accuracy, size of feature subsets. |
PDF Full Text Request