Title: An integrated model for long-term water consumption prediction

Authors: Fang He; Tianyu Zhao; Fanhao Kong

Addresses: School of Urban Construction, Wuhan University of Science and Technology, Wuhan 430065, China ' School of Urban Construction, Wuhan University of Science and Technology, Wuhan 430065, China ' School of Urban Construction, Wuhan University of Science and Technology, Wuhan 430065, China

Abstract: This article focuses on prediction model in long-term water consumption. The principal component analysis (PCA) was utilised to diminish dimensionalities. The Hodrick-Prescott (HP) filter decomposition technique was applied to perform PCA and separate the time series into their respective trend and cyclical series. The Grey Model (GM) was applied to predict the trend series of water consumption, while the Bayesian Optimised Least Squares Support Vector Machine (BOLSSVM) model was utilised to predict the cyclical series. Thus a PCA-GM-BOLSSVM model was constructed. Based on the water consumption data from 2008 to 2022 in Wuhan, the PCA-GM-BOLSSVM model reduced the mean relative and absolute errors by 28.71% and 27.82%, compared to the PCA-GM-LSSVM model. Compared to the ANN model, the PCA-GM-BOLSSVM model achieved an 80.63% and 80.33% reduction in mean relative and absolute errors. The PCA-GM-BOLSSVM model enhanced fitting accuracy and reduced error rates.

Keywords: water consumption prediction; principal component analysis; HP filter decomposition; GM-BOLSSVM model.

DOI: 10.1504/IJWMC.2026.155340

International Journal of Wireless and Mobile Computing, 2026 Vol.31 No.1, pp.79 - 92

Received: 09 Jan 2026
Accepted: 07 Mar 2026
Published online: 30 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article