Ground-level O-3 sensitivity analysis using support vector machine with radial basis function

摘要

Previous research studies have revealed human susceptibility to tropospheric ozone and consequently huge amount of investments allocating to monitor and research about this pollutant. High expenditures of monitoring the air contaminants and needing for spacious facilities can be decreased by applying the soft computing approaches and new technologies. In this paper, support vector machine, a powerful approach with reliable results in previous studies, is applied to predict the tropospheric ozone for Tehran, Iran, metropolitan area. Four photochemical precursors and three meteorological parameters are assumed as predictors. The impact of all parameters is scrutinized, and the best model of input combinations is depicted: RMSE=0.0774 and R=0.8459. Trying to find out each parameter impact, the datasets are divided into different groups and are used as input combinations where the outcomes depicted that particulate matters (PM2.5 and PM10), ambient air temperature (T), CO and NO2 are the most effective parameters on the O-3 value tolerances. Kernel trick as a crucial function for support vector machines is analyzed in the current study, and the radial basis function is illustrated as the best kernel trick for this study in comparison with polynomial, linear, multilayer perceptron tricks. Finally, to calibrate the measuring instruments, using the support vector machine with radial basis function can represent an acceptable result for the best input combination.