Research On The Relationship Between Syndromes And Drugs In The Treatment Of Diabetic Nephropathy In Traditional Chinese Medicine Based On Domain-driven Data Mining

This study compares and analyzes the shortcomings of traditional data-driven data mining,and it was considered that in the mining of TCM knowledge,domain knowledge and expert experience should be combined to change data-driven data mining into domain-driven data mining,and domain-related knowledge should be integrated under the guidance of synthetic method to optimize the construction of data mining model.This research mainly includes three parts.The first part is the research of domain-driven data mining model,including data preprocessing,selection of models and algorithms,evaluation and interpretation of mining results driven by domain knowledge.In the second part,the synthetic method is introduced to explores the Man-machine combination of Man-centered domain-driven data mining model.The third part takes the clinical dialectical treatment of diabetic nephropathy by traditional Chinese medicine as an example,discusses the relationship between syndrome and medicine of diabetic nephropathy from the perspective of domain-driven,and explores the law of medication of diabetic nephropathy by dialectical treatment of traditional Chinese medicine.1.Purpose1.1 Combining the characteristics of TCM data,integrating domain knowledge to optimize data mining model construction;1.2Construct a set of domain-driven data mining model under the guidance of synthetic method to explore the comprehensiveness,accuracy and availability of knowledge acquisition;1.3Taking Chinese medicine diagnosis and treatment of diabetic nephropathy as an example,under the guidance of the above theoretical research,systematically search for TCM clinical syndrome differentiation treatment of diabetic nephropathy,using the method of association rules to explore the syndrome-drug relationship of diabetic nephropathy and explore the drug rule of diabetic nephropathy,to provide theoretical and data support for the later construction of TCM clinical diagnosis and treatment decision-making database for diabetic nephropathy.2.methods and content2.1 System to retrieve the CNKI,Wanfang database,VIP database(the retrieval time is from the self-built library until December 2018)of the relevant literature of TCM clinical dialectical treatment of diabetic nephropathy,draw the basic information of the patients,and the doctor of traditional Chinese medicine diagnosis,syndrome,symptom,treatment,prescription,medicine and etc.,Establishing of a literature-based Chinese medicine diagnosis and treatment database for diabetic nephropathy;2.2Data preprocessing:combine the books of traditional Chinese medicine,and under the guidance of domain experts,perform data cleaning,data integration,data conversion and data reduction on the extracted data.2.3Data analysis and model construction:extract the target data set,use Apriori algorithm to conduct association rule analysis on drugs with different syndromes,obtain the number of rules generated under different support and confidence conditions by improving Apriori algorithm,and select the corresponding drug combination with fewer than 15 rules.2.4 Model evaluation and result interpretation:introduce the method of interest evaluation,and re-evaluate the mining results from the two aspects of the degree of promotion and usefulness of the rules.3.ResultsThis study analyzes the shortcomings of traditional data mining.Based on the complexity and multi-dimensional characteristics of traditional Chinese medicine data,under the guidance of the synthetic method,establish a Man-machine combination of Man-centered domain-driven data mining model,including data preprocessing,selection of models and algorithms,evaluation and interpretation of mining results,and added a feedback link in the output section to enrich and correct the domain knowledge base.In this study,a total of 310 literatures related to the dialectical treatment of diabetic nephropathy were included,and 442 TCM dialectical treatment schemes were extracted.Through the statistics of syndrome type and traditional Chinese medicine,there are 7 TCM syndromes mainly involved in clinical diabetic nephropathy.respectively is qi and Yin deficiency,liver-kidney deficiency,spleen-kidney deficiency,yin-yang deficiency,shi-zhuo deficiency,blood stasis deficiency and zhuo-du deficiency,Among them,Qi and Yin deficiency is the most common syndrome of diabetic nephropathy,a total of 177 treatments,accounted for 40%.A total of 208 Chinese herbal medicines were used in 442 programs,and the frequency of drug use was 4,844.There are 25 Chinese medicines used more than 50 times,the cumulative frequency of use is 3041 times,accounting for 62.77%of the total frequency of use.Statistical categories for medicine,found that tonify deficiency,heat,promoting blood circulation to remove blood stasis drug use frequency is higher,The top five traditional Chinese medicines with the highest frequency of use are Huangqi、Fuling、Shanyao、Danshen and Shanzhuyu.The association rules of 7 syndromes of diabetic nephropathy were analyzed,and the core drug combinations of 7 syndromes were obtained.The obj ective interest and subj ective interest of the obtained rules were evaluated.The core drug combinations under different syndromes of diabetic nephropathy were finally obtained by eliminating redundant rules and evaluating the promotion and usefulness of the rules.Four strong association rules were finally obtained for Yin and Yin deficiency syndrome,suggesting that the commonly used combination of drugs for diabetic nephropathy with Qi and Yin deficiency syndrome was the combination of Shengdihuang and Huangqi,Shanyao、Fuling and Huangqi,Shanyao、Shanzhuyu and Huangqi,Fuling、Shanzhuyu and Huangqi.The strong association rules of seven medication rules were finally obtained for the syndrome of liver and kidney deficiency,suggesting that the commonly used medication combinations of liver and kidney deficiency type diabetic nephropathy were Fuling、Shanzhuyu and Shanyao,Shengdihuang、Shanyao and Shanzhuyu,Mudanpi、Shanzhuyu and Shanyao,Zexie、Shanzhuyu and Shanyao,Mudanpi、Shanyao、Fuling and Shanzhuyu.The strong association rules of six medication rules were finally obtained for the deficiency of spleen and kidney syndrome,suggesting that the commonly used medication combinations for the deficiency of spleen and kidney syndrome of diabetic nephropathy are Danshen、Fuling and Huangqi,Filling、Dangshen and Huangqi,Baizhu、Dangshen and Huangqi,Danshen、Shanyao and Huangqi,Danshen、Baizhu and Huangqi,Danshen、Zexie and Huangqi.The strong association rules of six medication rules were finally obtained for Yin-Yang deficiency type,suggesting that the commonly used medication combinations of Yin-Yang deficiency type diabetic nephropathy are Shanyao、Zexie、Shanzhuyu and Fuling,Filling、Shanzhuyu、Mudanpi and Shanyao,Shanyao、Shanzhuyu、Mudanpi and Filling,Shanzhuyu、Shudi and Fuling,Huangqi、Zexie and Fulling,Fuzi、Shanzhuyu and Fulling.Five strong association rules were finally obtained for blood stasis syndrome,suggesting that the commonly used combination of drugs in blood stasis syndrome with diabetic nephropathy were Danshen.、Shanzhuyu and Huangqi,Yimucao and Huangqi,Shengdihuang、Shanzhuyu and Huangqi,Danshen、Danggui and Huangqi,Dahuang、Danshen and Huangqi.Three strong association rules of medication rules were obtained for the syndrome of dampness and turbidity,suggesting that the commonly used combination of drugs in dampness and turbidity with diabetic nephropathy were Shanyao、Baizhu and Fuling,Fuzi、Baizhu and Fuling,Shanyao、Zexie and Fuling.The strong association rules of seven medication rules were obtained for turbid toxin syndrome,suggesting that the commonly used medication combinations of turbid toxin syndrome with diabetic nephropathy were DanshenHuangqi、Zexie and Fuling.Danshen、Shanzhuyu and Huangqi,Danshen、Fuling、Baizhu and Huangqi,Shanyao、Fuling、Huangqi and Danshen,Fuling、Shanzhuyu and Huangqi,Danshen、Fuling、Shanzhuyu and Huangqi。4.ConclusionThe domain-driven data mining model based on integrated integration combines domain knowledge,experts and machines to construct a human-machine-based domain-driven data mining model.Under the guidance of the integrated integration method,domain-driven data mining can effectively reduce the number of redundant rules,and in the case of expert participation,more valuable and useful rules can be mined.