Multi-headed CNN-GRU model for particulate matter (PM2.5) concentration prediction in smart cities

Air pollution is becoming a major concern these days considering the increased number of vehicles on roads and industrialisation. This is creating a higher impact on human health. To deal with pollution levels and control it in smart city environment, predicting pollution level at a higher accuracy is very important. This will help monitor air quality and take measures to prevent pollution occurrence and avoid its effect. The objective of this work is to propose a novel multi-headed CNN-GRU model which has a higher accuracy. This model is comprising of multiple convolutional neural network (CNN) models for capturing the features of multiple variables of air pollutant concentration data. Information is then concatenated and transferred to the gated recurrent unit (GRU) layers and then to dense layer for single output as a next hour pollution concentration prediction. The model gives the best performance when compared to other deep learning models.