Water consumption prediction in hebron using RBF neural networks

Abstract

The lack of water in Palestine in general is a serious problem due to the geopolitical issues, as an urgent solution that will reduce the suffering of the citizen is to achieve a fair distribution for the available quantities of water; this requires a prediction of the water consumption for the Hebron citizen. So, our work aims to predict the amount of water consumption for a given customer in a given season. The prediction is based on the customer's data during the last two years (2007, 2008) from the database of Hebron municipality. In order to achieve the project goal, we implement a radial basis function network, to achieve acceptable prediction for future consumption of the water. Before using the input data, a normalization and processing of the data is carried out, to get the suitable input for the training phase of the neural network. The results show that radial basis function networks are efficient when used to solve prediction problems. Also, the results show that the training method is the most important part when building the neural network. In the case of our project we noticed that the training still needs much work to reduce the error and to permit the network to accept the available input data with high variance