Abstract

The quality of air we breathe has a considerable effect on our health\nand well-being. Environmental studies have confirmed a dramatic increase in the\nconcentration of harmful gases in the air in the last few years. Therefore,\ndetermining the conditions that trigger high concentration of these pollutants and\ntheir timely forecast is useful for a cleaner and healthier environment. The current\nwork proposes a real time model for predicting the concentration of Sulphur\nDioxide, Nitrogen Dioxide, Nitrogen Oxide, Carbon Monoxide and Ozone in the\nair based on the meteorological and temporal conditions analysis. It also analyzes\nthe inter-relationship and correlation between the presence of two pollutants in\nthe air. The prediction model is developed using Deep Belief Networks based on\nRestricted Boltzmann Machines. The observed accuracy of the model is 88.79%.\nSuch models can be useful for government to take preventive actions and frame policies which are conducive to environmental health and also assist the general\npublic in making decisions about their commute to locations. We have used the\ntime series data gathered from the year 2000 to 2012 for this model which\ncorresponds to 11 measuring stations at Attica Basin, Greece.