Using a Hybrid Model to Forecast the Prevalence of Schistosomiasis in Humans.

Lingling Zhou,Shunxiang Cai,Ying Wang,Shaofa Nie,Jing Xia,Lijing Yu,Yun Shi

doi:10.3390/ijerph13040355

Abstract

Background: We previously proposed a hybrid model combining both the autoregressive integrated moving average (ARIMA) and the nonlinear autoregressive neural network (NARNN) models in forecasting schistosomiasis. Our purpose in the current study was to forecast the annual prevalence of human schistosomiasis in Yangxin County, using our ARIMA-NARNN model, thereby further certifying the reliability of our hybrid model. Methods: We used the ARIMA, NARNN and ARIMA-NARNN models to fit and forecast the annual prevalence of schistosomiasis. The modeling time range included was the annual prevalence from 1956 to 2008 while the testing time range included was from 2009 to 2012. The mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were used to measure the model performance. We reconstructed the hybrid model to forecast the annual prevalence from 2013 to 2016. Results: The modeling and testing errors generated by the ARIMA-NARNN model were lower than those obtained from either the single ARIMA or NARNN models. The predicted annual prevalence from 2013 to 2016 demonstrated an initial decreasing trend, followed by an increase. Conclusions: The ARIMA-NARNN model can be well applied to analyze surveillance data for early warning systems for the control and elimination of schistosomiasis.

Highlights

Schistosomiasis is an acute and chronic, neglected tropical parasitic disease that is globally distributed in 78 countries, including Africa, Asia, the Middle East, and South America [1]
The Autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of different original prevalence series (OS) are displayed in Figure 1, the Figure 1A,B,E, and F which collectively suggest that the series was non-stationary
We found the minimum Bayesian information criterion (BIC) (5, 7) = 0.7895 and BIC (5, 7) = 0.6716, resulting in an order of auto-regression of p = 5 and the order of moving average of q = 7

Summary

Introduction

Schistosomiasis is an acute and chronic, neglected tropical parasitic disease that is globally distributed in 78 countries, including Africa, Asia, the Middle East, and South America [1]. 2013, while the actual number of treated people in that year was only 42.1 million. This great deficit underscores both the profound impact that schistosomiasis has on worldwide populations and the fact that it is often neglected. Our purpose in the current study was to forecast the annual prevalence of human schistosomiasis in Yangxin County, using our ARIMA-NARNN model, thereby further certifying the reliability of our hybrid model. Methods: We used the ARIMA, NARNN and ARIMA-NARNN models to fit and forecast the annual prevalence of schistosomiasis. The modeling time range included was the annual prevalence from 1956 to 2008 while the testing time range included was from 2009 to 2012. We reconstructed the hybrid model to forecast the annual prevalence from 2013 to 2016. Conclusions: The ARIMA-NARNN model can be well applied to analyze surveillance data for early warning systems for the control and elimination of schistosomiasis

Objectives

Methods

Results

Conclusion