Discuss Data Preprocessing And Data Mining Of Medical Records Of Distinguished Veteran Doctors Of TCM

Objective: Discuss data preprocessing and data mining of medical records of distinguished veteran doctors of TCM.Methods:1. Medical record data preprocessing of traditional Chinese medicine:(1) Data preprocessing of herbs: Standarding the standard name and alias of herbs, concluding four qi, five flavors, channel entry, classify, and effect.etc, establishing code of herbs, code of statistics, and setting up data table.(2) Data preprocessing of formula: Standarding the standard name and alias of formula, collecting classes, provenance, directions, action, indication, calcite.etc establishing code of formula, and setting up data table of formula..(3)Data preprocessing of symptoms:Screening descriptive words of symptoms, conducting classification and mergence, and setting up data table of similar symptoms.(4) Data preprocessing of efficacy evaluation: Referencing Guideline of New Traditional Chinese Medicine in Clinical Research, selecting and scoring cardinal symptoms, secondary symptoms, and setting up data table of efficacy evaluation.2.Medical record data mining of traditional Chinese medicine:(1)Analysis of drug-treated:(1) On the basis of previous data preprocessing, with the aid of the building medical record database of distinguished veteran doctors of TCM, building on the drug-treated frequency and resistance frequency of statistical analysis;(2) Through the medical record database of distinguished veteran doctors of TCM export data. By using SAS software, building of traditional Chinese medicine(TCM) Logistic regression and Varclus clustering method;(3) Through the medical record database of distinguished veteran doctors of TCM, built using Apriori algorithm for mining method of drug-treated combination.(2)The syndrome differentiation model based on BP neural network: With the aid of the building medical record database of distinguished veteran doctors of TCM, according to the binary classification method to establish the data table, with the help of Matlab software, to discuss the construction method of the syndrome differentiation model.(3) Analysis of “quantization”: For the nature and flavor, channel entry and effect of formula quantitative calculation, establish a quantitative method of “quantization”.(4) Based on the concept of "seed" information mining extraction method of symptoms: With the aid of ICTCLAS segmentation system for medical record word processing, screening of 24 seed concepts, scanning the medical records, to obtain the compound seed concept, symptom information.Results:1. Medical record data preprocessing of traditional Chinese medicine: The data preprocessing of 700 rich herbs,418 formulas,79 similar symptoms groups, 20 common diseases efficacy score have been completed, the medical record database of distinguished veteran doctors of TCM has been successfully imported, the input function of the medical record has been implemented.2. Medical record data mining of traditional Chinese medicine:(1) Analysis of drug-treated:(1) Analysis of drug-treated frequency: In case the corresponding medical records, the drug-treated frequency and resistance frequency in the corresponding medical records have been counted;(2) By Logistic regression analysis to get the general and specific index of drug-treated, through the Varclus clustering analysis, obtained the classification results of drug-treated.(3) Through the Apriori algorithm, dig out the drug combination and summarize empirical formula of distinguished veteran doctors of TCM.(2) The syndrome differentiation model based on BP neural network: qi vacuity pattern, for example, the syndrome differentiation model of qi vacuity pattern has been constructed successfully, the total accuracy rate reached 80%.(3) Analysis of “quantization”:Take Xuefu Zhuyu Decoction and Shaofu Zhuyu Decoction for examples, calculate the quantitative results of the two four qi, five flavors and channel entry, in accordance with the characteristics of clinical application.(4) Based on the concept of "seed" information mining extraction method of symptoms: 217 compound seed concepts have been successfully extracted.Conclusion:Effective data preprocessing of medical records of distinguished veteran doctors of TCM is a necessary condition for the construction of distinguished veteran doctors of TCM medical case database, and meanwhile brings convenience to data mining. Multi-angle and multi-level medical records of the distinguished veteran doctors of TCM medicine data mining, provides methods to explore the experience of distinguished veteran doctors of TCM, and makes better use of the medical records of distinguished veteran doctors of TCM and their experience in the clinical practice.