Application of SARIMA-RBF combination model in fitting and predicting the incidence of hand, foot, and mouth disease

Abstract

Objective To explore the application effect of SARIMA-RBF combination model in fitting and predicting the incidence of hand, foot, and mouth disease (HFMD). Methods The SARIMA model was established using the monthly incidences of HFMD from January, 2009 to June, 2015 in China. The fitted value based on the SARIMA model was taken as the input vector, and the actual value was used as the output vector. Two SARIMA-RBF combination models were established according to whether the time factor was used as the input vector (the model with the time factor was named as combination model A, the model without the time factor was named as combination model B). The predicted incidences of HFMD from July to December, 2015 by SARIMA model and the two combination models were compared with actual values, so as to evaluate the results of fitting and prediction by the models. Results The SARIMA (1, 0, 1) (0, 1, 1)12 model was the most appropriate one. In the fitting phase, the MAPE, MER, MSE and MAE fitted by the SARIMA model, combination model A and the combination model B were 19.985%, 16.177%, 8.608, 2.029; 16.661%, 11.940%, 3.741, 1.502 and 21.487%, 15.998%, 7.590, 2.013, respectively. The MAPE, MER, MSE and MAE predicted by the SARIMA model, combination model A and the combination model B were 9.119%, 8.988%, 1.874, 1.107; 6.536%, 7.395%, 1.926, 0.911 and 12.016%, 11.140%, 2.370, 1.372, respectively. Conclusions The SARIMA-RBF combination model with the time factor was superior to the SARIMA model in fitting and prediction, and could be used for short-term prediction of HFMD. Key words: SARIMA; RBF neural network; Combined model; Hand, foot, and mouth disease