Evaluation possibility of Particular Matter prediction by using Neural Network

Particular matters are one of the important air pollutants that have direct effects on human health. In this research, by comparing, feed forward ANN and NARX has been estimated particular matter of Tabriz city. metrology data and air quality data from 2013 to 2017 has been used. Particulate matter estimated by considering temperature, wind speed and rain precipitation in each model and the results compared. Also PM2.5 data form BaghShomal air quality station in Tabriz has been used. 50 present of data used for testing and validation and the rest of data used for training network The results showed that best state estimating with seasonal effect belong to feed forward ANN train with amounts of R=0.85, MSE=0.057and without seasonal effect belong to NARX with amounts of R=0.999, MSE=0.005. Modeling results with real data showed that best results belongs to feed forward ANN with 0.0007 error.