Using Support Vector Machines to Predict the Variation of Organic Pollutants in Pond Water

There is a growing perception that the research of fishery water quality forecasting are popular and important topic today, due to the food quality security and health impact caused by exposing to water pollutants existing in aquaculture. This paper aims to prove the feasibility of predicting the organic pollutant levels of pond water via SVM. The experimental data in weekly time series are collected from the fishery ponds in Xiao Tangshan in Beijing. The functional characteristics, including the network structure, the kernel function selection and the parameter sensitivity of SVM are investigated. The performance of the SVM model and the conventional BP neural network in predicting is also compared.