Temporal-Spatial Aggregated Urban Air Quality Inference with Heterogeneous Big Data

Recently air quality information has drawn much attention from public and researchers as deteriorated air quality extremely damages human health. Meanwhile the limiting number of air quality monitor stations and complexity of influencing factors on air quality raise the starving demand on future air quality prediction. In this paper we propose a temporal-spatial aggregated urban air quality inference framework using the heterogeneous temporal and spatial datasets to infer the future air quality. We deeply analyse the influencing factors on air quality in terms of temporal and spatial features and then elaborately design a linear regression-based inference model with offline parameters learning and real time predicting. We not only estimate the parameters for our model itself, but also estimate the correlation parameters of single factor on the air quality in order that the model can make prediction on future air quality precisely. Based on real data sources, we appraise our approach with extensive experiments in Beijing and Suzhou. The results show that with the superior parameters learning, our model overmatches a series of state-of-art and commonly used approaches.