This document discusses using an artificial neural network to forecast electricity demand. It describes preprocessing data, creating a feed-forward neural network model with input, hidden and output layers, and training the model using backpropagation and incremental training. The model is trained on 80% of the data and tested on the remaining 20%. Mean square error is used to evaluate accuracy on both the training and test sets, with a lower error on the test set indicating better generalization of the model to new data. The goal is to accurately forecast future electricity demand based on input variables like population, GDP, price indexes, and past consumption data.