The Study Of Yunqi Theory Periods Based On Periodogram

ObjectiveThe study of yunqi doctrine has always been an important part of TCM research.For a long time,researchers has never stopped researching the doctrine using meteorological data and clinical data.After the founding of new China,the meteorological observatory network throughout the country.Profit from the accumulation of meteorological data,researchs of the doctrine were carried out across the country,began in 1980s.Methodology,also gradually tend to objective,from the enumeration method to mathematical statistics.However,the evaluation criteria is confused and the use of indicators is different,on the current studies of the doctrine.For what kind of indicators can reflect Host Evolutive Phase(HEP),Host Climatic Qi(HCQ),Year Evolutive Phase(YEP)and Year Climatic Qi(YEQ)still lack of systematic research.In the discussion of the periods of yunqi doctrine,the vast majority of the current researches are based on the discussion in the time domain,while ignoring the fact that the periods of the doctrine is frequencies.In view of this,the purpose of this study is to make a systematic study of the meteorological indicators to discuss the doctrine.And using the ground temperature data and clinical cases of pneumonia to discuss the periodicity of the doctrine,based on the frequency domain.MethodsStatistical analysis and plotting are processed by R.3.3.2.Re-encode each year of data the Great Cold is the beginning of yun-qi-anniversary,and the last day of the Slight Cold is the end of yun-qi-anniversary.With the year in which most of the time is the mark of yun-qi-anniversary.Then the structures of yun-qi,from 1952 to 2014,were calculated.The study is divided into four parts:In the first part,the correlation between temperatures and the periods of yun-qi doctrine was dicussed.the means of daily temperatures of Guangzhou and Beijing which was grouped by Mounth-Day,HEP-Year,YEP-Year and HCQ-Year,YCQ-Year respectively,were calculated.The groups,HEP-Year,YEP-Year,HCQ-Year and YCQ-Year,were tested by One-way ANOVA.The groups,Mounth-Date,HEP-Year,YEP-Year,HCQ-Year and YCQ-Year,were fitted by cosine line model which alse was used to predict the temperatures in 2012-2014.Derivatives of Mounth-Date groups were calculated.The means of YEP-Year,YCQ-Year were grouped by YEP,YCQ,YEP period and YCQ period respectively,and tested by One－way ANOVA.Tukey’s test was used to compare the two.Guangzhou 60 years temperatures was ranked,the first and the last 5%were named the coldest and the hottest weather.They were grouped by YEP,YCQ and tested by Chi-square test.Periodogram was used to estimate the periods of the frequencies of the coldest and the hottest weather.And we chose 10-year periods and 6-year periods from cryptic periods and discussed the relationship between the periods and YEP,YCQ.In the second part,the extraction and normalization of traditional Chinese medicine symptoms was dicussed.Diagnostics,Diagnostics of Chinese Medicine and 1000 marked pneumonia medical records were used to establish three symptoms normalization lexicons:symptoms normalization lexicon,symptoms description lexicon and keywords-adjectives lexicon.Symptoms-description were pre-process by regular expression.Based on the maximum probability method,Chinese word Segmentation Program(CSP),Direct Extraction Program(DEP)and Combination Extraction Program(CEP)were built.These three programs were compared by the counts of dimensions,the manual processing and the efficiency of symptoms extraction.In the third part,the dimensions reduction of traditional Chinese medicine symptoms was dicussed.Normalization symptoms dataset was used.Compared the efficacy of dimensions reduction between different neurons numbers in 5 layer DBA(Deep Belief network Aotoencoder).Compared the efficacy of dimensions reduction between 5-layers DBA,3-layers DBA and LPCA(Logistics PCA).In the fourth part,the cluster of pneumonia cases and its correlation about HEP and HCQ was dicussed.Gender distribution about HEP and HCQ was tested by Chi-square test.Age distribution about HEP and HCQ was tested by Kruskal-Wallis H test.LPCA.16 was used to reduce dimensions.The scores,computed by LPCA.16,of each cases were clustered by K-means.Each cluster was described by 10 highest constituent ratio symptoms and frequent item sets which found by association rules analysis.Chi-square test was used to discuss the correlation between pneumonia clusters and HEP,HCQ.Periodogram was used to found out HEP-period and HCQ-period.The dynamic relationship between clusters and HEP,HCQ was described by cosine lines.ResultsFor the first part,there is no significant difference among the groups YEP-Year,YCQ-Year,YEP-HEP,YCQ-HCQ,according to One-way ANOVA.The models of cosine line work fitted well(P<0.0001,adj R2>0.95)in the groups Mounth-day,HEP,HCQ and the true values of 2012-2014 also in 95%CI of the predictions(R2>0.95).Based on Tukey’s test,the last two periods were hotter than the first two periods in both YEP-period and YCQ-period.Years in Yang had more coldest weather than years in Yin.Tai-Yu,Shao-jiao,Tai－Zheng,Shao-Gong,Tai-Shang had more hottest weather than the others.Shao-Yin-Jun-Huo,Shao-Yang-Xiang-Huo,Tai-Yang-Han-Shui had mode coldest weather,while Shao-Yin-Jun-Huo,Tai-Yin-Shi-Tu,Shao-Yang-Xiang-Huo had mode hottest weather.Chi-square test showed that the distribution of the coldest and the hottest weather were different between YEP or YCQ(P<0.0001).The coldest weather had 2-year-period,3-year-period,5-year-period.6-year-period,10-year-period were exist though combind the periods.The hottest weather had 6-year-period,20-year-period.For the second part,compared with CSP,CEP and DEP can reduce the dimensions.But DEP had lots of symptoms descriptions needed to be processed by manual.CEP could reduce about a half of the manual processing.Two programs had a good effect of symptom extraction.DEP had higher precision rate.But CEP was more effect by raising the recall rate.The diff of counts of symptoms descriptions was reduced in the form of an index.For the third part,under the current sample size,the effects of reducing dimension were decreased with the increase in the number of neurons in the second layer,On 5-layer DBA.Compared with 3-layer DBA and LPCA,5-layer DBA was most effect,when the number of dimensions was lass than 10,.Otherwise,LPCA was most effect.Some of the pattern of LPCA.16 were similar to TCM syndrome.For the fourth part,Chi-square test showed that the distribution of gender between YEP and YCQ,were no significant difference.So did the distribution of ages,by testing in Kruskal-Wallis H test.Ward Hierarchical Clustering showed that the cases could be dividen into 5 clusters.After cluster the cases by K-means method,5 clusters had different performance.Cluster 1 was cough,white sputum,shortness of breath,anorexia,less sputum,poor sleep,chest tightness,sparse sputum,poor spirit,chest pain.Cluster 2 was cough,shortness of breath,white sputum,anorexia,phlegm,sleep difference,yellow sputum,poor spirit,fever,chest tightness.Cluster 3 was cough,fever,aversion to cold,anorexia,headache,yellow sputum,dizziness,poor spirit,fatigue,less sputum.Cluster 4 was cough,fever,aversion to cold,less sputum,white sputum,dry mouth,anorexia,shortness of breath,sore throat,yellow sputum.Cluster 5 was fever,anorexia,poor spirit,poor sleep,aversion to cold,dry mouth,fatigue,cough,dizziness,shortness of breath.The distribution of each cluster in frequencies were related to HEP and HCQ.Periodogram showed that all clusters had HEP-period and HCQ-period except cluster 5 in HCQ-period.Conclusion1.The means of tempreture from different groups are all mainly performed HEP,HCQ effect.Cosine line model can be used to fit the mean temperature of date,HEP and HCQ.The models can reflect some characteristics of HEP and the HCQ,such as:the characteristics of stabilities,regionalism change,climatic change and period.2.The distribution of the coldest and hottest weather have significant differences between YEP,YCQ.6-year-period and 10-year-period are exist in both two kinds of weather.But the performances of period are different.The hottest weather have hold 6-year-period and 10-year-period,while coldest weather should combine 2-year-period,3-year-period,5-year-period.3.CEP and DEP based on the maximum probability method is efficient on symptoms extraction.4.In the current sample size,if the output dimensions is small,it is appropriate to use 5 layer DBA,otherwise,should use LPCA.5.The distributions of each cluster of pneumonia cases are related to HEP and HCQ.HEP-period and HCQ-period are widely exist in different clusters.6.The dynamic relationship between clusters and HEP,HCQ was described by cosine lines.7.Meteorological indicators,syndrome frequency distribution of the time series,exist but nit limit the periods of yunqi doctrine,the doctrine should be used flexibly.