Water Consumption Prediction of Semarang Water Utilities using Support Vector Regression Radial Basic Function Kernel Method

Abstract

People have various needs that can’t be released considering their role as living things. The diversity of human needs requires planning for the future, one of which is the provision of water supply because water needs are increasing. Estimation models can be done using the Support Vector Regression (SVR) method. SVR is a development of the Support Vector Machine for regression cases. SVR has four kernels that are commonly used, and in this study, the kernel used is the Radial Basis Function Kernel because RBF is considered capable of maintaining good predictive accuracy. The purpose of this study is to apply the SVR method to predict water consumption with the Radial Basis Function kernel by getting the best SVR parameter and knowing the error value generated from the SVR method. The data used in this study is Semarang Water Utilities’ (PDAM) water consumption data from January 2013 to March 2018. The SVR method test results obtained the best parameters are lambda (λ) = 10, sigma (σ) = 0.001, cLR = 0.01, C (Complexity) = 0.01, epsilon (ɛ) = 0.00000001, with the number of iterations = 1000, produces the lowest Mean Absolute Percentage Error (MAPE) is 1.751%.