Formula Mining Of Chronic Kidney Disease Based On Weighted Association Rule In Traditional Chinese Medicine

Based on long-term clinical practice,traditional Chinese medicine(TCM)has developed a unique theoretical framework and a diagnosis and treatment model that has played an important role for safeguarding people's physical and mental health.TCM prescriptions(i.e.,formulas),which are the primary carriers of TCM for treating diseases,are the organic combination of one or more herbs.Based on the patient's illness and the characteristics of herbs,these prescriptions are formulated according to the TCM compatibility theory.Compatibility,a critical factor for enhancing the clinical efficacy of TCM prescriptions,is a rule that governs the organic combination and coordinated application of selected herbs in formulas.Currently,chronic kidney disease(CKD)has become a global health problem;however,TCM is extremely beneficial for alleviating symptoms in CKD patients.Association rule mining can effectively mine the compatibility patterns from prescriptions.These patterns provide an objective reference for prescriptions and assist physicians to design optimal compatibility of herbs in prescriptions for treating CKD.Moreover,such patterns can improve the efficacy of herbs and natural drugs for new drug development and discovery,which is of great importance for modernizing TCM.In this thesis,the compatibility patterns of CKD are mined based on therapeutic effects.The primary contributions of this thesis are summarized below.(1)To improve the efficiency of mining TCM compatibility for treating CKD,we proposed a high-compression tree structure(HCPT-tree)and an efficient frequency pattern mining algorithm(HCPT-growth)to mine CKD compatibility patterns and obtain core drugs;moreover,drug pairs and drug combinations are commonly used in the TCM treatment of CKD.Using path counting node,HCPT-tree supports more prefix sharing and significantly reduces the number of tree nodes.Therefore,the HCPT-growth algorithm can reduce the number of recursive calls and conditional trees via HCPT-tree.Experimental results indicate that HCPT-growth has a quick run speed,requires less memory,and demonstrates better scalability compared to the existing algorithms such as LP-growth,NSFI,and MAFIA-FI.(2)Efficacy is the impetus and foundation of the development of TCM.However,existing mining of compatibility patterns only considered the support in the prescription without considering the efficacy,and some herbs with high occurrence frequency do not have good efficacy.Thus,some herbs with good efficacy but low occurrence freq uency may be ignored.We propose a weighted association rule algorithm,called WHCPT,to mine compatibility patterns with good efficacy for CKD.The WHCPT algorithm is applied to mine the weighted frequent patterns to obtain the core herbs,couplet herbs,and herb combinations with the best efficacy for treating CKD.Clinical trials demonstrated that these herbs and its combinations can improve the clinical symptoms of CKD patients.(3)The minimum weighted support threshold ? must be determined prior to mining association rules.However,this support threshold is difficult to determine.Setting a low ? produces a large number of short patterns.Conversely,setting a high ? might lead to missing some useful long patterns.Currently,determination of ? is primarily based on repeated experiments,which is inefficient.We propose a TOP-K weighted association rule algorithm,which we refer to as PWFP-growth,to mine compatibility patterns with the best efficacy for treating CKD.The proposed algorithm only requires setting the pattern length and number of patterns to be mined.It does not require setting the minimum weighted support threshold ?.PWFP-growth can estimate ? when constructing the global tree,which has dramatically improved algorithm efficiency.Experimental results demonstrate that PWFP-growth has higher efficiency compared to the existing algorithms,such as CRMN and BOMO.(4)TCM prescriptions in clinical and medical research are increasing continually and dynamically.However,existing researches on the compatibility patterns of TCM prescriptions have not adopted the dynamic mining method.We propose an incremental TOP-K weighted association rule algorithm called IWFP-growth that can continuously mine compatibility patterns in the increasing number of prescriptions in a dynamic prescription database.When the prescriptions used in clinical trials are added to the dynamic prescription database,IWFP-growth algorithm results are updated to reflect the added information regarding the herbs and its combinations with good efficacy in treating CKD.These herb combinations can help TCM physicians continuously adjust the clinical medication selection and design the proper prescription to treat CKD.