Forecasting PM2.5 concentration using artificial neural network and its health effects in Ahvaz, Iran

The main objective of the present study was to predict the associated health endpoint of PM2.5 using an artificial neural network (ANN). The neural network used in this work contains a hidden layer with 27 neurons, an input layer with 8 parameters, and an output layer. First, the artificial neural network was implemented with 80% of data for training then with 90% of data for training. The value of R for the data validation of these two networks was 0.80 and 0.83 respectively. The World Health Organization AirQ(+) software was utilized for assessing Health effects of PM2.5 levels. The mean PM2.5 over the 9-year study period was 63.27(mu g/m(3)), about six times higher than the WHO guideline. However, the PM2.5 concentration in the last year decreased by about 25% compared to the first year, which is statistically significant (P-value = 0.0048). This reduced pollutant concentration led to a decrease in the number of deaths from 1785 in 2008 to 1059 in 2016. Moreover, a positive correlation was found between PM2.5 concentration and temperature and wind speed. Considering the importance of predicting PM(2.5 )concentration for accurate and timely decisions as well as the accuracy of the artificial neural network used in this study, the artificial neural network can be utilized as an effective instrument to reduce health and economic effects.

Automated summary

The study aimed to predict the health endpoint of PM2.5 using an artificial neural network. Positive correlation was found between PM2.5 concentration and temperature and wind speed. The reduced pollutant concentration led to a decrease in the number of deaths, indicating the potential of the artificial neural network in reducing health and economic effects.