Ambient Temperature Forecasting Using Non-linear Autoregressive Neural Networks Anas Kabbori
Abstract
This paper aims to forecast ambient temperature values of Marrakesh City, using non-linear autoregressive neural networks. The process of forecasting ambient temperature, provides valuable data that improves prediction and prevention of natural disasters, especially heat waves and drought. Relying on Non-linear Auto-Regressive (NAR) neural networks and Non-linear Auto-Regressive with eXogenous input (NARX) neural networks on a dataset containing 12 years (2007-2019) of temperature data. The resulting values showed good performance using both networks. NARX networks showed better results in training and in testing phases. Those findings can be a stepping stone in agricultural seasons planning or modeling natural hazards and preventing them.