Application of multiple seasonal ARIMA model in predication of birth defect incidence in Xi'an area / 西安交通大学学报(医学版)

ABSTRACT

Objective To predict the incidence of birth defects in Xi'an using the auto-regressive integrated moving average product seasonal model.Methods In Xi'an,the trend of the incidence of birth defects was analyzed and tested from October 2009 to August 2015.Using the data from September to December 2015,the actual birth defects were compared with the model fitting data to evaluate the predictive performance of the model.Multiple seasonal ARIMA model was then fitted under time series to predict the incidence of birth defects in 2016.Results Seasonal effect was seen in the incidence of birth defects in Xi'an.A multiple seasonal ARIMA(0,0,1) (0,1,1)12 was established.The mean of absolute error and the relative error were 9.5 and 0.084,respectively,when compared to the simulated number of patients from September to December in 2015,suggesting that ARIMA (0,0,1) (0,1,1)12 has a better predictive ability.Results under the prediction of multiple seasonal ARIMA model showed that the number of patients in 2016 was similar to that of 2015 in Xi'an,with a slight increase and a decrease in the peak value.Conclusion Multiple seasonal ARIMA(0,0,1)(0,1,1)12 model could be used to successfully predict the incidence of birth defects in Xi'an.