Prediction of Cholera Incidence by Using the Comparison of Four Models: Autoregressive Integrated Moving Average Model, Holt Model, Brown Model and Simple Regression Model

Abstract

Aims: The forecast is a topical subject, which aid in decision making and its performance. The aim of this study is to predict the disease between 1995 and 2010. Place and Duration of Study: The choice of the disease is of after its appearance in our survey in the region of Gharb. Time series were illustrated between1988-1994. Regional cholera annual data reported from ministry of health of Morocco. Methods: The comparison of four models by the analysis of the series of cholera cases includes examining graphic series by using EVIEWS software, the consideration of the autocorrelation and partial autocorrelation functions, define the model that suits, estimate it, diagnose, the residue analysis and compare the four models to choose the best for use in the forecasting process. Except the stationary series, we used IBMSPSS V22 for the other steps. Results: Throughout this work, it is assumed that the underlying structure of the series follows an autoregressive integrated moving average (ARIMA) process. It is presumed that observations of the disease follow an autoregressive moving average process of order AR (1) and therefore ARIMA (1, 1, 0). The comparison of models of time series is extended away by using the statistics fit of the model: MAPE, BIC and R-squared, in addition to the sig. of the parameters and the analysis of residues by Ljung-Box and Durbinwatson statistic. The validation of the series is estimated by the calculation of the Mean Absolute Percentage Error (MAPE) and the signification of the parameter with P =0,05. Conclusion: Brown model is the model of choice for the prediction of cholera cases.