Forecasting Fine-Grained Air Quality for Locations without Monitoring Stations Based on a Hybrid Predictor with Spatial-Temporal Attention Based Network

Air pollution in cities is a severe and worrying problem because it causes threats to economic development health. Furthermore, with the of industry technology, rapid population growth, massive expansion cities, total amount emissions continue increase. Hence, observing predicting air quality index (AQI), which measures fatal pollutants humans, has become more critical recent years. However, there are insufficient monitoring stations for AQI observation construction maintenance costs too high. In addition, finding an available suitable place high density difficult. This study proposes spatial-temporal model predict long-term city without stations. Our calculates correlation between station region using attention mechanism leverages distance information all existing target regions enhance effectiveness structure. we design hybrid predictor that can effectively combine time-dependent time-independent predictors dynamic weighted sum. Finally, experimental results show proposed outperforms baseline models. ablation confirms structures.