Cikapundung streamflow forecast modeling using support vector machine

Abstract

As regard to its key role in provision of raw water for Bandung City, Cikapundug River has been suffering from issues such as water pollution, water conflict, decline of stream flow, et cetera. Facts showed that rapid urbanization and land use change have critical impacts on water availability within the basin. In order to be able to estimate the amount of water available in the Cikapundung River, this study make use of SVM (Support Vector Machine), to forecast the stream flow. The forecasting process is based on 4 scenarios with giving different length of data, starting from 3 years up to 6 years of monthly stream flow data. By considering 2 independent variables, i.e. rainfall and evapotranspiration, the analysis is done to identify the influence of data length on forecasted flow and evaluate its accuracy compared to observed flow. The results show that 4 years of data length give least deviation, with 20.94% error in total, and 4 years data could represent the better model in Upstream Cikapundung River.