Method and Grid Workflow on Water Consumption Forecasting of Ziyang City Based on Support Vector Machine Trained by Genetic Algorithm

Hao-Lan      Yang; Shi-Chang      Li; Feng-Chuan      Zhang; Zhong-Fu      Wu

doi:10.2174/187221210790244712

Abstract

Accurate forecasting of urban water consumption can provide guidance to urban water supply and planning. In recent patents, artificial neural network (ANN) has been applied in urban water consumption forecasting. However, the practicability of artificial neural network is affected due to the weaknesses, such as over-fitting, slow convergence velocity and local extremum. Support vector machine (SVM) is a novel machine learning method, which has excellent generalization ability in the situation of small sample. In SVM, the choice of training parameters has an important influence on the forecasting performance of SVM. Thus, support vector machine trained by genetic algorithm is proposed to forecast urban water consumption in the study, in which genetic algorithm is used to determine the training parameters of SVM and improve the forecasting performance of SVM. Water consumption of Ziyang city in China from 1998 to 2007 is used to study the forecasting performance of the proposed model. The experimental results indicate that the proposed model can gain higher forecasting accuracy than grey model, artificial neural network.

Keywords: Support vector regression, Workflow, Training parameters, Water consumption, Ziyang city, Forecasting model