Možnosti zpřesňování predikce koncentrací znečišťujících látek v ovzduší pomocí neuronových sítí

Abstract

The work builds on the results of the thesis Markéta Dřezová and focuses on the possibility of more accurate predictions of air pollutants, specifically PM10. Were used type of neural network multilayer perceptron network and Kohonen maps. For input parameters were used pollutant concentrations measured from the previous day's air pollution monitoring station in Ostrava - Poruba and meteorological parameters measured at the meteorological station in Mošnov in the years 2005 - 2009.