Forecasting daily pollen concentrations using data-driven modeling methods in Thessaloniki, Greece

摘要: Abstract Airborne pollen have been associated with allergic symptoms in sensitized individuals, having a direct impact on the overall quality of life considerable fraction population. Therefore, forecasting elevated airborne concentrations and communicating this piece information to public are key issues prophylaxis safeguarding In study, we adopt data-oriented approach order develop operational models (1–7 days ahead) daily average highly allergenic taxa: Poaceae, Oleaceae Urticaceae. The developed using representative dataset consisting meteorological time-series recorded during years 1987–2002, city Thessaloniki, Greece. input variables (features) optimized by making use genetic algorithms, whereas evaluate performance three algorithms: i) multi-Layer Perceptron, ii) support vector regression iii) regression trees originating from distinct domains Computational Intelligence (CI), compare resulting traditional multiple linear models. Results show superiority CI methods, especially when several ahead, compared Furthermore, complement each other, combined model that performs better than one separately. ranges, terms index agreement, 0.85 0.93 clearly suggesting potential latter ones can be utilized provision personalized on-time services, which improve citizens.