Application Research Of Osteoarthritis Chinese Medicine Molecular Analysis Based On SVM

Osteoarthritis is a chronic joint disease that occurs in the elderly population, and Chinese medicine treatment which has the whole and local governance, both specimens, less cost, few side effects and other advantages is one of the main means of therapy of osteoarthritis. Using computer simulation technology and system biology technology and others to analyze the molecules that have drug activity in the prescription of traditional Chinese medicine can promote the progress that developing new drugs of osteoarthritis according to traditional Chinese medicine.Support Vector Machine technology in pattern recognition based on the solid statistical learning theory has been widely applied to the design and analysis of drug and other related fields of medicine. This paper applies binary-class and single-class based on SVM to a series of studies of osteoarthritis compound Chinese medicine, which focus to build a related SVM model with the guidance of QSAR theory to analyze the property of the molecular in traditional Chinese medicine for osteoarthritis. The paper focuses on the research of optimizing the performance of SVM model and applies the results to the analysis of Chinese medicine for osteoarthritis. The main tasks of the thesis are as follow:(1) The paper systematically reviews Statistical Learning Theory, the principle of the single-class and the binary-class in the SVM algorithm, the properties of the kernel function, and studies the training algorithms of SVM.(2)After studying the attribute reduction in rough set, we use it as an attribute selector in the pre-process of the SVM model and get the more typically minimal reduction of attribute-set, reduce the space dimensions of the training samples. Finally above it we establish the SVM model and compare it with the model which doesn’t have a attribute selector, we find that the former has a better performance to eliminate the noise disturbance and its structure is more simple.(3)The considered problems in the building of the SVM model include:the choices of the selection of kernel function and how to set kernel parameters. To consider the learning and the generalization performance of SVM at the same time, we apply the hybrid kernel to SVM, and the related experiments with the date sets of UCI shows that the performance of the hybrid kernel is better than the single kernel function’s. As for kernel parameters optimizations, the paper combines the cross-validation strategy with the grid search method to globally search for some kernel parmeters.lt overcomes some flaws from the artificial search such as one sidedness of experience and the low time efficiency. Finally, we integrate these methods into SVM model to do drug analysis for the traditional Chinese medicine and the molecules in it in terms of osteoarthritis.(4)It analyzes the traditional Chinese medicine for the osteoarthritis from a macroscopic perspective. Under the guidance of the theory of Chinese herbal property and effect, the paper uses the improved SVM algorithm to build a effect-classifier which uses the properties, flavors, channel from traditional Chinese medicine as sample input variables and aims to identify the efficacy of traditional Chinese medicine. After the efficacy of traditional Chinese medicine from the Osteoarthritis Compound is clear, then we analyze the biological property of the molecule in these medicine through two experiments.Firstly,we reveal that the molecules in Chinese herbs have various properties of drug activities with one class classification approach which is known as data description. Then we create target classifiers with the improved SVM algorithm to classify the activity of Chinese medicine molecular and analyze their drug activities and reveal their pharmacological effects on osteoarthritis.Under Windows environment the subject uses VC++6.0 to develop a visualization SVM analysis software based on the VC version of LIBSVM, and the software can be applied to the problems such as single-class, binary-class, multi-category, regression and so on. It provides an effective attempt to identify the efficacy of traditional Chinese medicines with the computer software and reveal the multi-target characteristics of traditional Chinese medicine molecular.