Forecasting of ozone concentration in smart city using deep learning

Clean air is one of the most important needs for the well-being of human being health. In smart cities, timely and precise air pollution levels knowledge is vital for the successful setup of smart pollution systems. Recently, pollution and weather data in smart city have been bursting, and we have truly got into the era of big data. Ozone is considered as one of the most air pollutants with hurtful impact to human health. Existing methods used to predict the level of ozone uses shallow pollution prediction models and are still unsatisfactory in their accuracy to be used in many real-world applications. In order to increase the accuracy of prediction models we come up with the concept of using deep architecture models tested on big pollution and weather data. In this paper, a new deep learning-based ozone level prediction model is proposed, which considers the pollution and weather correlations integrally. This deep learning model is used to learn ozone level features, and it is trained using a grid search technique. A deep architecture model is utilized to represent ozone level features for prediction. Moreover, experiments demonstrate that the proposed method for ozone level prediction has superior performance. The outcome of this study can be helpful in predicting the ozone level pollution in Aarhus city as a model of smart cities for improving accuracy of ozone forecasting tools.