Développement d’algorithmes pour l’analyse des séries temporelles des données de production d’eau potable.

Abstract

This research work is interested in the modeling of time series, it is to analyze and model stoochastic optimization problems and the propose a forecast model applied to the production of drinking water.We used the monthly drinking water consumption data for the city of Biskra. The objectif of work is to devlop an algorithm to study the techniques of analysis and modeling of empirical time séries corresponding to the methodology of Box and Jenkins to generate an ARMA model (autoregressive and moving average models) Of drinking water demand in the city of Biskra which is to allow a prediction of the future evolution of a phenomenon and used for the validation of the production values of water seized monthly using tools of the main components used during the identification phase are therefore the plots of the series, autocorrelation correlograms (FAC), and autocorrelation partial (FACP). The modeling allowed us to determine how many autoregressive parameters (p) and moving average (q) are needed to obtain an effective and parsimonious model of the process. The analysis of the ARMA residuals constitutes an important test of the model,the information criteria AIC and BIC are proposed to test the quality of the chosen model. The proposed method is applied to water resources production dataset..