Contribution à la prédiction des précipitations par l'apprentissage automatique (machine learning)

Abstract

The evolution of climate conditions and their impact on water resources present new challenges for the management of hydraulic infrastructures, especially wastewater networks. In this context, precipitation forecasting becomes crucial. This study explores the application of Long Short-Term Memory (LSTM) neural networks to predict rainfall over a period of 120 months. The goal is to enhance the anticipation of extreme rainfall events and improve the planning of sewage systems. The results demonstrate the effectiveness of the LSTM model in predicting hydrological time series with a good fit between real and predicted values