科研之友
微信
新浪微博
Facebook
分享链接
摘要
Accurate and reliable prediction of diarrhoea outpatient visits is necessary for the health authorities to ensure the appropriate action for the control of the outbreak. In this study, a novel method based on time series decomposition and multi-local predictor fusion has been proposed to predict the diarrhoea outpatient visits. For time series decomposition, the Ensemble Empirical Mode Decomposition with Adaptive Noise (EEMDAN) is used to decompose diarrhoea outpatient visits time series into a finite set of Intrinsic Mode IMF) components and a residue. The IMF components and residue are modeled and predicted respectively by means of Generalized Regression Neural Network (GRNN) as local predictor. Then the prediction results of all components are fusioned using another independent GRNN as fusion predictor to obtain final prediction results. This is the first study on using a EEMDAN and GRNN to constructing an prediction model for diarrhoea outpatient visits prediction problems. The pre-procession and post-processing techniques are used to take into account the seasonal and trend effects in the datasets for improving the prediction precision of proposed model. The performance of the proposed EEMDAN-GRNN model has been compared with Seasonal Auto-Regressive Moving Average (SARIMA), Single GRNN, Wavelet-GRNN and also with EEMD-GRNN by applying them to predict four real world diarrhoea outpatient visits. The results indicate that the proposed EEMDAN-GRNN model provides more accurate prediction results compared to the other traditional techniques. Thus EEMDAN-GRNN can be an alternate tool to facilitate the prediction of diarrhoea outpatient visits.
