Deep Multi-Task Learning Based Urban Air Quality Index(AQI) Forecasting And AQI Level Estimation

Air pollution affects the ecological environment and human health.Obtaining comprehensive and accurate air quality information is conducive to people’s arrangements for daily life for protecting people’s health from air pollutants.Limited urban air quality monitoring stations cannot provide air quality in areas without monitoring stations and air quality in the future.Urban air quality is influenced by many complex factors,e.g.,meteorology,traffic,factory pollution emission,POIs,and road network.Existing urban air quality estimation and prediction methods do not adequately consider information related to air quality,e.g.,factory pollution emission.They define and extract the features manually which are often over-specified,incomplete and take a long time to design and validate.In addition,existing urban air quality estimation models often cannot achieve good generalization performance due to the lack of training samples.This paper proposes a deep multi-task learning based urban AQI forecasting and AQI level estimation method(DAFE).On one hand,a variety of air quality-related urban big data information(meteorology,traffic,factory pollution emission,POIs,and road network,etc.)are considered,and deep neural networks and graph embedding methods are used to learn the representation of these relevant sequential and static information as well as build the correlation between the air quality and these representation.Thus the AQI level of the regions without monitoring stations can be estimated with the air quality of the monitoring stations and the AQI in the future of the monitoring stations can be forecasted.On the other hand,the multi-task learning is used to solve the AQI forecasting and AQI level estimation tasks jointly,which can improve the generalization performance of single task model through share the representation between related tasks.In this paper,a large number of experiments are conducted based on the dataset of Hangzhou City.The experimental results show that the proposed method is superior to the existing air quality forecasting and estimation methods.