Fuzzy system models combined with nonlinear regression for daily ground-level ozone predictions

Characterization of the Concentrations of Volatile Organic Compounds in the Romanian Littoral using General Regression Neural Networks: A Case Study

“…Designing models for pollutant dispersion in crowded zones is a complex process, due to the configuration of antropic and natural structures (e.g., urban constructions, forestry fund, and water), of their different dynamic and radiative properties (Kalabokas et al 2001), and of the evolution of atmospheric factors that must be modeled by appropriate techniques . Starting in 1990, a variety of models were developed for pollutant dissipation, including Gaussian processes for environmental impact assessments and ozone concentration evaluation (Grasic, Mlakar, and Boznar 2006;Olesen, Berkowicz, and Lofstrom 2007;Petelin, Grancharova, and Kocijan 2013); linear simple and multiple models for pollutant correlation in the Romanian Littoral ; artificial neural networks for forecasting the concentration of airborne pollutants in urban areas and modeling the dissipation of inorganic pollutants in urban and industrial zones (Barbes et al 2009; Barbulescu and Barbes 2013); fuzzy systems for forecasting the ozone concentration (Lin and Cobourn 2007;Nebot, Mugica, and Escobet 2008); and generalized autoregressive linear models for particulate dispersion (Hrdlickova et al 2008). Also, different integrated modeling systems for simulation and prediction of atmospheric pollution dispersion (Karppinen et al 2000;Owen et al 2000;Carruthers, Sheng, and Vanvyve 2008;Macintosh et al 2010;Zou et al 2010) or for air quality simulation have been constructed (Paatero 2000;Matthias 2008;Pay et al 2010;Appel et al 2011).…”

Forecasting peak air pollution levels using NARX models

“…Notwithstanding these important advantages, the class of polynomial NARX models has not been employed before, to the authors' knowledge, for ground-level ozone forecasting. More precisely, nonlinear regression models have been presented in Cobourn (2007) and Lin and Cobourn (2007) in the Kentucky metropolitan area, using as input a nonlinear function, but this approach did not include the autoregressive part of the NARX models. Also, in the cited study there was no automatic selection of the best nonlinear terms to reconstruct ozone concentrations.…”

Hybrid Fuzzy Auto-Regressive Integrated Moving Average (FARIMAH) Model for Forecasting the Foreign Exchange Markets

“…(6). Then the minimal fuzziness of these parameters is determined using the same criterion as in the Eq.…”

“…Andres et al [5] proposed a strategy for constructing a hybrid model, which combines the fuzzy clustering and the multivariate adaptive regression splines (MARS) in order to use their theoretical advantages of these models for bankruptcy forecasting, especially when the information applied for forecasting is drawn from company financial statements. Lin and Cobourn [6] combined TakagiSugeno fuzzy system and a nonlinear regression (NLR) model for daily ground-level ozone predictions. Chang et al [7] developed a hybrid model by integrating fuzzy rule base (FRB), self-organization maps (SOMs), and Genetic Algorithms (GAs) to forecast the future sales of a printed circuit board factory.…”