ARTIFICIAL NEURAL NETWORK MODELING OF PM10 AND PM2.5 IN A TROPICAL CLIMATE REGION: SAN FRANCISCO DE CAMPECHE, MEXICO
In this paper, a computational methodology based on Artificial Neural Networks (ANN) was developed to estimate the index of PM10 and PM2.5 concentration in air of San Francisco de Campeche city. A three layer ANN architecture was trained using an experimental database composed by days of the week, time of day, ambient temperature, atmospheric pressure, wind speed, wind direction, relative humidity, and solar radiation. The best ANN architecture, composed by 30 neurons in hidden layer, was obtained using the Levenberg-Marquardt (LM) optimization algorithm, logarithmic sigmoid and linear transfer functions. Model results generate predictions with a determination coefficient of 93.01% and 90.10% for PM2.5 and PM10, respectively. The proposed methodology can be implemented in several studies as public health, environmental studies, urban development, and degradation of historical monuments.