A hybrid prediction model applied to diarrhea time series

Accurate and reliable prediction incidence of diarrhea disease is necessary for the health authorities to ensure the appropriate action for the control of the outbreak. In this study, a hybrid prediction algorithm (EMD-GRNN), which combines empirical mode decomposition (EMD) as time series decomposition method and the generalized regression neural network (GRNN) as prediction model, is proposed to improve the quality of diarrhea prediction. First, the proposed EMD-GRNN algorithm decomposes the complex original diarrhea time series into a series of intrinsic mode functions (IMFs) and a residual series. The IMF components and residual series are than applied training datasets to model and predicted using GRNN model. Finally, these local prediction results are combined into the final diarrhea prediction result using another independent GRNN model by a trainable combinatorial method. The proposed EMD-GRNN algorithm is evaluated by predicting the diarrhea cases number of children and adult located in Shanghai of China. The obtained experimental results confirm that the proposed EMDGRNN algorithm is better than the traditional autoregressive integrated moving average (ARIMA) model and the single GRNN model.