Application of a support vector regression on prediction of bacillary dysentery combined with meteorological and air pollutants index / 中华疾病控制杂志

ABSTRACT

Objective To explore the application of support vector regression (SVR) model combined with meteorological and air pollutants index in the prediction of the cases of bacillary dysentery in Lanzhou City, so as to provide scientific reference for the prevention and control of bacillary dysentery.Methods Time series data of the reported cases of bacillary dysentery from December 2013 to August 2016, combined with the meteorological and air pollutants data, were used as training set to fit support vector regression model. The data from September 2016 to December 2017 was used as validation set to verify the model and compare the effect in fit and prediction with different models. Results A total of 7 192 bacillary dysentery cases were reported in Lanzhou City from 2013 to 2017. The correlation coefficient of meteorological and pollution factors with the cases of bacillary dysentery was more than 0.4, except air pressure. The parameters of the fit model were selected based on the integrated data, acquiring the three parameters with the smallest test error were C=5, γ=0.02 and ε=0.000 1, respectively. The validation set was used to test the different models, which showed that the integrated data model had the best predictive accuracy and robustness . The root mean squared error (RMSE) was 0.164 7 and the mean absolute percentage error (MAPE) was 16.405%. Conclusion SVR model combined with meteorological and air pollutants index is effective in the prediction of bacterial dysentery.