Ecological Study And Forecast Of Tuberculosis And Meteorological Factors In Shanxi Province Based On Dynamic Bayesian Network Model

Objective:Study the spatiotemporal aggregation characteristics of tuberculosis monitoring data in 11 prefecture-level cities in Shanxi Province from 2008 to 2017,explore the meteorological factors that may affect tuberculosis,and use meteorological factors to establish an early warning model based on the dynamic Bayesian network(DBN),Compared with the static Bayesian(BN)and support vector machine models,the classification and recognition performance of the three models is evaluated,and finally it provides scientific and rational basis for the decision-making of tuberculosis prevention and control in Shanxi Province.Methods:In this study,the epidemiological characteristics of tuberculosis in Shanxi Province were described using tuberculosis surveillance data and synchronic meteorological data from 11 prefecture-level cities in Shanxi Province from 2008 to 2017;cluster analysis of multi-indicator panel data was used to explore the spatial clustering of tuberculosis;The moving percentile method is used to classify the risk level of tuberculosis,so as to study its time-gathering characteristics;use cross-correlation analysis to determine the lag period of meteorological factors for tuberculosis;through Bayesian structure learning and parameter learning,a dynamic shellfish is finally established Yesh Network Early Warning Model.Results:1.From 2008 to 2017,the reported incidence of tuberculosis in Shanxi Province generally showed a downward trend.Each year,it showed the phenomenon of middle high and two low,with certain seasonal characteristics.The number of reported cases was the lowest in January-February and March was the highest in the whole year.Peak,with significant seasonal increase in spring and early summer;the average reported incidence over a ten-year period is from high to low.The top five prefecture-level cities are Datong(69.38 / 100,000),Xinzhou(66.14 / 100,000),and Yuncheng.Cities(65.60 / 100,000),Luliang(60.95 / 100,000),Linfen(60.31 / 100,000),all of which are higher than the ten-year average of the incidence of tuberculosis in Shanxi Province(55.67 / 100,000);multi-indicator panel data The results of the cluster analysis of 11 divided the 11 cities into two regions in order to model and analyze for different regions.The first area includes Taiyuan,Changzhi,Jincheng,Jinzhong,and Luliang,and the second area includes Datong,Shuozhou,Yuncheng,Xinzhou,and Linfen.2.Based on the results of cross-correlation analysis,it is reasonable to use dynamic Bayesian model fitting with meteorological factors lagging for 2 months;the monthly average temperature and monthly precipitation are positively correlated with the incidence of tuberculosis.The mean air pressure has a negative correlation with the incidence of tuberculosis,and the rest of the regions have no obvious rules to follow.After principal component analysis,the results show that seven meteorological factors such as monthly rainfall are the main factors affecting the incidence of tuberculosis in Shanxi Province.3.The classification and recognition performance comparison of the three models found that the classification accuracy of the DBN model in the two regions was the highest,the first region was 95.00%,the second region was 97.50%,and the remaining precision,TPR,TNR,and F-Measure,G-measure and other indicators are also the highest DBN.Among them,the DBN has the largest F-measure value,and the two regions are 0.77 and 0.84,which indicates that DBN reflects the performance of a few classes better than the other two models;the G-measure value is the largest,which is 0.86 and 0.85 respectively,indicating that The comprehensive classification performance of the minority class and the majority class is better;the AUC of the two regions in order from large to small is DBN> BN> support vector machine.Conclusion:1.Shanxi Province has tuberculosis clustering in time,space and time.Peak incidence in spring and early summer.The average reported incidence in ten years from high to low.The top five prefecture-level cities are Datong,Xinzhou,Yuncheng,Luliang,Linfen.Seven meteorological factors such as monthly precipitation are the main factors affecting the incidence of tuberculosis in Shanxi Province.2.For panel data with two dimensions of space and time,the classification and identification performance of the dynamic Bayesian network early warning model of tuberculosis-meteorological factors established in this study is significantly better than the static Bayesian network and support vector machine model.DBN can accumulate the law and experience of variable changes over time,and combine the newly obtained information to better predict the future moment.The results of this study provide new ideas and methods for Shanxi Province's tuberculosis prevention and control decisions.