Research On Automatic Discovery Of Derivation Relationship Of Analogous Formulae

Chinese medical formulae (hereinafter referred to as formulae) are combination of drugs for treating diseases, which are guided by Chinese medicine theory and according to certain compatibility principles. A large number of classic formulae, selected and accumulated by clinical practice in thousands of years, are carrier of treatment experiences and wisdoms of doctors in past dynasties.Analogous formulae, characterized by similar composition and treatment efficacies, consist of a earlier-created root formula and some derived formulae. The research on analogous formulae is significant to improve the formulae theory and clinical practices. On the one hand, analogous formulae as formulae system developed from traditional classic formulae, are the essence of ancient doctors’application of classic formulae, and the guideline of clinical formulae medication including drug addition and subtraction. The research on the analogous formulae is also the process to learn derivation rules of ancient formulae. On the other hand, analogous formulae is a shortcut to learn pharmacology of traditional Chinese medical formulae. Once the root formulae and variation rules are mastered, clinical problems can be easily coped with. Furthermore, the similarities on compositions and efficacies of analogous formulae are also very import to uncover the regulation on making up the formula.A large number of classic formulae were selected and accumulated by clinical practice in thousands of years, not to mention those new formulae created in clinical practice constantly. Mistakes are unavoidable if analogous formulae are collected and analyzed manually. Based on the computer technology, the automatic analysis and sorting of the analogous formulae will greatly increase the efficiency of research. So, the research on derivation rules, mining connotation of analogous formulae derivation, building its mathematical model, and realizing automatic reorganization can not only help build the analogous formulae knowledge system, but also promote the study and application of formulae.Based review of literature, there is little research on automatic reorganization of analogous formulae derivation. The difficulties on this topic can be summed up as follows:1) Connotation of analogous formulas is fuzzy, and has not reached a consensus. How to modeling the analogous formulas derivation?2) The model of analogous formulas derivation is related to formulas similarities. However, existing similarity models cannot satisfy the requirement of modeling analogous formulas derivation. A comprehensive model need to be developed based on formula’s origin and development, nature, flavor, meridian, efficacy similarities and so on.3) The process of analogous formulas derivation is complicated. The model should not only contain multiple features, but also reflect how the formula’s efficacies changing with its drug compositions and dosages. That means the algorithm of formula’s efficacies calculation need be realized firstly, and then the measurement of efficacy changes between derived and source formula be implemented.According to solve aforementioned problems, this research implemented contents as follows:1) Chapter 1 introduced existing researches on concept and connotation of analogous formulas derivation, as well as other related topic including formulas similarity model, efficacy semantic network and algorithm of formula’s efficacies calculation, etc.2) Chapter 2 analyzed background and significant of this research and put forward research objectives.3) Chapter 3 carried on thorough analysis and formal representation on connotation of analogous formulas, analogous formulas derivation, analogous tree and other related concepts.4) Chapter 4 reported how basic data used in the research was collected, standardized, and stored. As a result, Chinese medicine database and formulas database were designed and built.5) Chapter 5 figured out a method to solve the measurement of closeness between drugs and their efficacies.6) Chapter 6 built a sophisticated semantic network of efficacies, which consists 218 nodes and 1405 edges.7) Chapter 7 designed and realized an algorithm of formula’s efficacies calculation based on the semantic network of efficacies. The test result is satisfied and the algorithm can calculate and represent efficacies of formulas more accurately and vividly.8) Chapter 8 proposed a method based on graph theory to measure the semantic distance of efficacies between different formulas. The measurement is proper to reflect efficacy changes as the formula derived from source formula.9) Chapter 9 executed an machine-learning experiment for analogous formulas derivation, which produced a model based on Logistic Regression.10) Chapter 10:The feasibility and practicality of the analogous formulas derivation model is verified through an application case:analogous formulas tree.11) Chapter 11 systematically summed up the research and put forward recommendations of future work.Innovations of this research are:1) built a sophisticated semantic network of efficacies for the first time, which include 6 semantic types between efficacies.2) realized an algorithm of formula’s efficacies calculation based on the semantic network of efficacies.3) proposed a method based on graph theory to measure the semantic distance of efficacies between different formulas.4) creatively established an analogous formulas derivation model based on Logistic Regression.