城市中的空氣汙染一直是個嚴重且受到關注的議題，同時嚴重的損害全世界人們的健康及生活，隨著科技工業的發展、快速的人口成長以及城市的快速擴張，越來越多的空氣汙染物因人們的活動而增加，在常見的幾種空氣汙染物中，PM2.5和人們的生活關係最大。因此為了人們的經濟發展以及身體健康，觀測及預測整個城市中的空氣品質指標(AQI)在近年來是個重要的研究議題。但在城市中並沒有密集的空氣品質監測站來達到這個目的，由於建置和維護監測站的成本過於昂貴，並且在已開發及人口密度高的城市裡，也難以找到足夠且空閒的土地用來建置。
因此在我們的論文裡，提出了一個時空間的深度學習模型來解決此問題，並對城市中沒有建立監測站地點的區域進行AQI的長期預測。我們的方法利用所有現有的監測站和目標預測區域之間的距離作為增強注意力架構學習的因子，並利用注意力機制來計算兩者之間的時空間相關性。此外我們更將預測值分解成兩個獨立且互補的分量，並設計一複合式預測架構分別對這兩個分量進行預測，最後再利用動態加權總合的方式讓這兩個獨立的預測分量，以更合理的方式作結合得到我們對沒有監測站地點未來AQI的預測值。

Air pollution in cities is a severe and concerning problem, meanwhile, it threatens seriously human health and life in the world. With the increasing development of industry and technology, the rapid growth of the population, and the massive expansion of cities, more and more pollutants are emit. There are several common air pollutants: O3、PM2.5、PM10、SO2、CO & NO2. Generally, PM2.5 is the most critical to Air Quality Index (AQI) and at the same time is fatal to human. Hence, observation and prediction of AQI in the whole city is important in recent years not only for economic development but human health. However, there are insufficient air quality monitoring station in a city to observe the AQI in the whole city. The spending to construct a station and maintain the operation is high, moreover, it is difficult to have an available and suitable place for monitoring station in cities with high population density.
This paper proposed a spatial-temporal model to predict the AQI in the whole city by making a long-term AQI prediction to the region without monitoring stations in city. Our model take advantage of distance between all existing monitoring stations and target region to enhance the attention structure, and calculating the spatial-temporal correlation of them by attention mechanism. Additionally, we divide predicted values into two independent and complementary portion, and design our hybrid predictor with two predictor separately to make prediction of two portions. In the end, dynamic weighted sum is exploited to sensibly combine the two independent portion.

[1] A. Appice, A. Ciampi, F. Fumarola, and D. Malerba. “Missing Sensor Data Interpolation. In: Data Mining Techniques in Sensor Networks,” SpringerBriefs in Computer Science, 2014. Springer, London. https://doi.org/10.1007/978-1-4471-5454-9_3
[2] L. Chen, Y. Ding, D. Lyu, X. Liu, and H. Long, “Deep multi-task learning based urban air quality index modelling,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 3, no. 1, pp. 1–17, 2019.
[3] W. Cheng, Y. Shen, Y. Zhu, and L. Huang, “A neural attention model for urban air quality inference: Learning the weights of monitoring stations,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 32, no. 1, 2018.
[4] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” 1997.
[5] H.-P. Hsieh, S.-D. Lin, and Y. Zheng, “Inferring air quality for station location recommendation based on urban big data,” in Proc. 21th ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, 2015, pp. 437–446.
[6] K. Kumar, A. Yadav, M. Singh, H. Hassan, and V. Jain, “Forecasting daily maximum surface ozone concentrations in Brunei Darussalam—an ARIMA modeling approach,” Journal of the Air & Waste Management Association, vol. 54, no. 7, pp. 809–814, 2004.
[7] M. Lateb, R.N. Meroney, M. Yataghene, H. Fellouah, F. Saleh, and M. Boufadel, “On the use of numerical modelling for near-field pollutant dispersion in urban environments- a review,” Environmental Pollution, vol. 208, pp. 271–283, 2016.
[8] V.-D. Le, T.-C. Bui, and S.-K. Cha, ‘‘Spatiotemporal deep learning model for citywide air pollution interpolation and prediction,’’ inProc. IEEE Int.Conf. Big Data Smart Comput. (BigComp), Feb. 2020, pp. 55–62.
[9] L. Li, X. Zhang, J. Holt, J. Tian, and R. Piltner, “Spatiotemporal Interpolation Methods for Air Pollution Exposure,” Symposium on Abstraction, Reformulation, and Approximation, 2011.
[10] Y. Lin, N. Mago, Y. Gao, Y. Li, Y.-Y. Chiang, C. Shahabi, and J. L. Ambite, “Exploiting spatiotemporal patterns for accurate air quality forecasting using deep learning,” Paper presented at the Proceedings of the 26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2018.
[11] Y. Lin, Y. Chiang, F. Pan, D. Stripelis, J. L. Ambite, S. P. Eckel, and R. Habre, “Mining public datasets for modeling intra-city PM2.5 concentrations at a fine spatial resolution,” InProceeding of the 25th ACMSIGSPATIAL International Conference on Advances in Geographic InformationSystems, 2017.
[12] L. Menut, B. Bessagnet, G. Siour, S. Mailler, R. Pennel, and A. Cholakian, ”Impact of lock-down measures to combat Covid-19 on air quality over western Europe,” Sci. Total En-viron. 741, 2020.
[13] Y. Qi, Q. Li, H. Karimian, and D. Liu, ‘‘A hybrid model for spatiotemporal forecasting of PM2.5based on graph convolutional neural network and longshort-term memory,’’ Sci. Total Environ., vol. 664, pp. 1–10, May 2019.
[14] Z. Qi, T. Wang, G. Song, W. Hu, X. Li, and Z. M. Zhang, “Deep air learning: Interpolation, prediction, and feature analysis of fine-grained air quality,” IEEE Trans. Knowl. Data Eng., vol. 30, no. 12,pp. 2285–2297, Dec. 2018.
[15] S. Qin, F. Liu, C. Wang, Y. Song, and J. Qu, “Spatial-temporal analysis and projection of extreme particulate matter (PM10 and PM2:5) levels using association rules: A case study of the Jing-Jin-Ji region, China,” Atmos. Environ., vol. 120, pp. 339-350, Nov. 2015.Le, Bui, et al. , 2020.
[16] Y. Rybarczyk and R. Zalakeviciute, “Machine learning approaches for outdoor air quality modelling: A systematic review,” Applied Sciences, vol. 8, no. 12, p. 2570, 2018.
[17] P.-W. Soh, J.-W. Chang, and J.-W. Huang, “Adaptive Deep Learning-Based Air Quality Prediction Model Using the Most Relevant Spatial-Temporal Relations,” IEEE Access, 6, 38186-38199, doi:10.1109/access.2018.2849820, 2018.
[18] N. Srivastava, E. Mansimov, and R. Salakhutdinov, “Unsupervised Learning of Video Representations using LSTMs,” 2015.
[19] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A.N. Gomez, L. Kaiser and I. Polosukhin, Attention is all you need. arXiv preprint arXiv:1706.03762 (2017).
[20] H. Wang, “Air pollution and meteorological data in Beijing 2016-2017,” 2019. Retrieved from: https://doi.org/10.7910/DVN/RGWV8X
[21] W.S. Wong, L. Yuan, and S. A. Perlin, “Comparison of spatial interpolation methods for the estimation of air quality data,” Journal of Exposure Analysis and Environmental Epidemiology, Vol 14, 404–415, 2014.
[22] T. Xayasouk, H.-M. Lee, and G. Lee, “Air Pollution Prediction Using Long Short-Term Memory (LSTM) and Deep Autoencoder (DAE) Models,” Sustainability, 12(6), 2570, 2020. Retrieved from https://www.mdpi.com/2071-1050/12/6/2570
[23] Y.-F. Xing, Y.-H. Xu, M.-H. Shi, and Y.-X. Lian, “The impact of PM2.5 on the human respiratory system,” 2016.
[24] X. Zhao, T. Xu, Y. Fu, E. Chen, and H. Guo, “Incorporating spatiotemporal smoothness for air quality inference,” in 2017 IEEE International Conference on Data Mining (ICDM). IEEE, pp. 1177–1182, 2017.
[25] Y. Zheng, X. Yi, M. Li, R. Li, Z. Shan, E. Chang, and T. Li, “Forecasting Fine-grained air quality based on big data,” in Proc. 21th ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, New York, NY, USA, pp. 2267-2276, 2015.
[26] Y. Zheng, F. Liu, and H.-P. Hsieh, “U-air: When urban air quality inference meets big data,” In Proceedings of the 19th ACM international conference on Knowledge discovery and data mining. 1436–1444, 2013.