Deep Convolutional Neural Network for Air Quality Prediction

Abstract

In this paper, we tackle air quality forecasting by using deep learning approaches to predict the hourly concentration of air pollutants (e.g., ozone, particle matter PM2.5 and sulfur dioxide). Deep learning (DL), as one of the most popular techniques, is able to efficiently train a scalable model on big data by optimization algorithms. The model is trained for air quality prediction with time series data. Our method takes the deep convolutional neural network (CNN) as the sequence module and inputs the time series data into the CNN model in turn for training. CNN is composed of many functional layers, such as convolution, pooling and ReLU. Convolution layer can effectively extract the sequential features of time series data. Sequential features work better than general features of time series data. Down-sampling in CNN is performed by the Pooling layer. Experimental results show that CNN performs well for air quality prediction.