Based On Machine Learning Technology, "treatise On The Side Permit The Analysis Method

This project combines the basic clinical research of traditional Chinese medicine and the research and development of the digitalization of the traditional Chinese medicine. It is based on the research on the method of analysis and mining of the famous old Traditional-Chinese-Medicine doctors'clinical experience which is a branch of the technical attack plan during the China's tenth five-year plan named data-mining-based research on the famous old traditional Chinese-medicine doctors'clinical treatment and dissemination method.This project is aimed at exploring the effective technique and method in the clinical experience of the famous old traditional-Chinese-medicine doctors. The research is based on machine-learning techniques and its content is the analytical method applied in Typhus Theory.On the basis of the theories of traditional Chinese medicine, the mathematical modeling and scientific computation of the internal attributes of traditional Chinese-medicine's prescriptions are roughly accomplished by means of information technology, artificial intelligence and mathematical principles. And on the basis of this, the classification and recognition experiment of the prescriptions and syndromes in Typhus Theory is carried out by means of the technique of the support vector machine.From the aspects of literature, theory, technique and method, software manufacture and its application, the machine-learning methods of prescription and syndrome in Typhus Theory are explored in a systematic way. And this provides a new insight for the quantification and classification of the prescriptions and syndromes in traditional Chinese medicine and a dependable method and technique for the mining and dissemination of the famous old traditional-Chinese-medicine doctors'clinical experience.1 Literature researchThe literature concerning the domestic and foreign research method and machine-le-arning technique of Typhus Theory is explored and organized in a systematic way.The research methods, particularly the mathematical methods, on Typhus Theory are summarized and reviewed in a detailed way. The research on the machine-learning methods both abroad and at home is also reviewed. Therefore, the current research situation is made explicit.2 Theoretical researchUnder the guidance of the basic theories of traditional Chinese medicine, and with a close reference to the traditional classification method of Typhus Theory and the current research situation of traditional Chinese medicine, the classification method of prescription and syndrome in Typhus Theory and the quantification method of traditional Chinese medicine knowledge are first described in a systematic way. It is stated that it will be a breakthrough for the quantification and digitalization of traditional Chinese medicine theories to seek an objective and commonly valuable method of quantification and classification of the prescription and syndrome in traditional Chinese medicine. Only in this way can the traditional Chinese medicine theories maintain its own characteristics and keep up with the times at the same time.3 Technical theoriy and its applications3.1 Approximate inference technique and the developing of Chinese medicine prescription intelligent analysis system(CPI AS).3.1.1 Approximate inference techniqueApproximate inference is inference based on uncertain knowledge and evidences. It is in fact a process of inferring from uncertain initial evidences and uncertain knowledge is used. The conclusion is basically reasonable but remains some uncertainty. In the knowledge and thinking behavior of human beings, certainty is relative while uncertainty is absolute. In the study of traditional Chinese medicine, the solution for the same result is not unique, and knowledge needed is not enough. Background knowledge possessed by experts is not always in agreement with each other. Therefore, the approximate inference method is needed in the computation process of the intelligent analysis system.Knowledge database is based on the review of literature concerned and the analysis of other people's research of Typhus Theory. It mainly includes the databases of medicine effects, features of effects and evidences of effects.3.1.2 The developing of Chinese Medicine Prescription Intelligent Analysis System(CPIAS).This project explores the medical principle of the CPIAS from the point of view of the starting point of the research and the aims of the CPIAS. The internal structure and the core of the system are designed. The main functions and the methods of realization of the system are also introduced in the end.The quality of medicine varies, resulting in the medicinal materials and the great scope of its application. It is hard to determine the quantity of medicine in prescription. There is no reference of comparison. Under the guidance of traditional Chinese medicine theories, a specific method of computation is designed to unite the quantity of medicine prescription. The concept of relative quantity is introduced. A model of computation of the relative quantity is proposed, thus providing a reference for the comparison of different intensity of medicine effects. At the same time, in the computation of the interaction intensity of two medicine effects, the concept of approximacy degree is introduced. Special computation models of the approximacy degree and the effect intensity of the medicine are designed. It can increase the weight in the computation of the relations of medicine effects, making the result more accurate than before. Through the interference of approximate inference and previous knowledge, the application evidences of prescriptions can be foreseen. Then, the most appropriate effect intensity can be obtained from the clinical evidences. After that, the application evidences of prescription is foreseen again and in the meantime the order of clinical manifestations(symptoms) is determinded.3.2 Support Vector Machines(SVM)and the developing of CMSVM machine-learning systemSupport Vector Machines based on solid theories, is an original study method of small samples. It does not concern the definitions of probability and mathematical laws. It has evaded the traditional process of induction and deduction. It has achieved the"transformational inference"from drilling samples to foreseeing samples. On the other hand, the ultimate determining function of the SVM is determined by a small number of support vectors. The complication of computation is determined by the number of support vectors instead of the dimensions of sample space. In some sense, this has escaped the problem of"dimension disasters". Due to the introduction of the core function, the classification model of the learning machine based on the SVM is a non-linear model and SVM can deal with non-linear classification effectively. The SVM is supported by statistical theories and the models created based on SVM can be transported easily. It can provide the transportability of the learning results and the boundaries of errors. The above-mentioned advantages of the SVM are suitable for the solution of the multi-dimensions, small-samples and non-linearity of the traditional Chinese medicine.This project explores the theoretical basis of the SVM and the classification principle of the SVM. A research team of CMSVM development is established and a machine-learning modeling system is designed. In the end the basic functions of the system are introduced.4 Research of software applicationsBy means of the CPIAS, this project makes a comprehensive quantitative analysis of the 195 prescription and syndrome which are chose from the Typhus Theory and made in agreement with each other, including the computation of relative medicine quantity, medicine quality intensity, medicine effect intensity, smell intensity and medicine Guijing intensity. Through the macro-quantitative analysis and the interference of knowledge and rules, the evidences of prescriptions are foreseen and are ordered by intensity. According to the clinical manifestations of the current evidences, the most appropriate effect intensity can be obtained based on the previous analysis. Based on the most appropriate effect intensity obtained, the evidences of prescriptions are foreseen again and also ordered by intensity. Based on the most appropriate effect intensity obtained and through the inference based on the interference of knowledge and rules, the order of clinical manifestations(symptoms) that the prescription concerns is determined with an analysis accuracy rate of 95%.After a quantitative analysis of the Typhus Theory through the CMSVM machine-learning modeling system, by means of the Mercer Kernel Function and the"intersection testing 1"and"the application of foreseeing"which are suitable for small samples, a drilling model can be established through eight methods and main prescriptions. The data obtained can be classified and foreseen effectively. There is research indicating that in the condition of small samples and if the quality of data obtained can be guaranteed the modeling has a high transportability no matter how many dimensions there are.The CPIAS and the CMSVM machine learning system can achieve the integrity of data and evade the influence of subjective factors. The dimensions of data are not reduced and the integrity of data is maintained. By applying the mathematical theories to the field of traditional Chinese medicine, the problems of multi-dimensions and non-linearity are solved and in the meantime combining the study method which are based on previous knowledge and the study method which are totally based on data, thus solving the problem of knowledge mining in the field of traditional Chinese medicine which has its uniqueness and commonalities with others.This project has proposed the computational method of prescription factors that is based on approximate reasoning theory and the research program of modeling inference that is based on the support vector method. And it has successfully carried out the quantitative analysis and classification and recognition of the 195 prescriptions and syndromes in Typhus Theory. The result of this research has provided a universal model for the research on the academic thoughts and experience of famous old traditional-Chinese-medicine doctors. In this research, the advanced machine learning theories and methods and traditional Chinese medicine are combined, thus solving the problem of the quantitative computation and classification of the relevant factors in the traditional Chinese medicine prescriptions and providing a solid basis for the future exploration of the deep knowledge and rules in the traditional Chinese medicine.