隨著城市化和工業化的嚴重，空氣品質逐漸下降。都市圈是受到空氣污染影響的最大的族群。隨著城市化和工業化的發展，空氣污染已經成為一個嚴重的環境問題。空氣品質預測被認為是重要的議題，早期預警和控制可以減少對都市居民的健康影響。本論文提出多通道時間混合神經網絡來預測空氣品質。該模型包括擴張卷積層、雙向閘循環單元層和注意力層，分別提取空間特徵、時間特徵和關鍵特徵。該混合了學習數據的不同維度，即時間維度和特徵維度。這結構該結構能夠建立時間特徵和特徵關聯，同時實驗證實此結構能夠準確預測空氣品質。我們在實驗中採用了來自（台灣 左營）、（中國 北京）、（哥倫比亞 甘迺迪）和（哥倫比亞 波利瓦爾公園）的數據集。實驗證明，我們提出的模型在大多數情況下都是贏過傳統的回歸方法、神經網絡和其他作者提出的方法。

In recent years increasingly serious air pollution, the air quality decline gradually. Especially people living in the city circle is affected by air pollution the most ethnic groups. With the quick advancement of urbanization and industrialization, air pollution has become a serious issue in developing countries. Air quality prediction is considered as a major issue, and early warning and control can reduce the health impact on the residents in the area. This paper aims to predict air quality by multi-channel temporal hybrid neural network. This model includes dilated convolution layer, bidirectional gate recurrent unit layer and attention layer which extract spatial features, temporal features and key features respectively. The hybrid structure learns different dimensions of data, namely the temporal dimension and the feature dimension. This allows the structure to build temporal dependencies and feature associations to learn higher-level features. The proposed model takes into account the first 168 hours of different pollutant data, meteorological data from a single station. We employ real world dataset from (Zuoying, Taiwan), (Beijing, China), (Kennedy, Colombia) and (Simon Bolivar Park, Colombia) in experiment. The results show that our proposed model is superior to the traditional regression method, neural network and other methods in most cases. It also shows that our proposed model can handle air quality predictions with high accuracy for next 24 hours ahead prediction.

Contents
Verification Letter i
Acknowledgement ii
Chinese Abstract iii
Abstract iv
List of Figures vii
List of Tables ix
Chapter 1 Introduction 1
Chapter 2 Related Work 4
2.1. Feature Selection 4
2.2. The Methods for Air Quality Prediction 4
2.3. Impact of Meteorological Conditions on Air Pollution 5
Chapter 3 Proposed Method 6
3.1. Normalization 7
3.2. Selection of Input Variables 8
3.3. Building Training Examples 9
3.4. Temporal Hybrid Neural Network 11
3.4.1. One Dimensional Dilated Convolution Layer 12
3.4.2. Bidirectional Gated Recurrent Unit Layer 14
3.4.3. Attention Layer 16
3.4.4. Dense Layer 17
Chapter 4 Experiments 19
4.1. Introduction to Datasets 19
4.2. Data Exploration 21
4.2.1. Missing Values 22
4.2.2. Data Distribution 23
4.3. Details of Experiments 25
4.3.1. Machine Learning Approach 26
4.3.2. Deep Learning Approach 26
4.4. Experiment Results 29
4.4.1. The result of feature selection 31
4.4.2. Comparison of Single Step Ahead Prediction Performance 33
4.4.3. Comparison of Multi-Step Ahead Prediction Performance 35
4.5. Statistical Analysis for Experiment Results 44
4.6. Discussion 45
4.6.1. The analysis of window size and structure 46
4.6.2. The Analysis of Machine Learning Method and Deep Learning Method 48
Chapter 5 Conclusion 53
Reference 54

Reference
[1] A. Bernhardt, "World’s worst pollution problems: the toxics beneath our feet," Zurich, 53p, 2016.
[2] P. J. Landrigan et al., "The Lancet Commission on pollution and health," The lancet, vol. 391, no. 10119, pp. 462–512-462–512, 2018.
[3] C.-H. Tseng et al., "The relationship between air pollution and lung cancer in nonsmokers in Taiwan," Journal of Thoracic Oncology, vol. 14, no. 5, pp. 784–792-784–792, 2019.
[4] M. Renzi, F. Forastiere, J. Schwartz, M. Davoli, P. Michelozzi, and M. Stafoggia, "Long-Term PM 10 Exposure and Cause-Specific Mortality in the Latium Region (Italy): A Difference-in-Differences Approach," Environmental health perspectives, vol. 127, no. 6, pp. 067004-067004, 2019.
[5] C.-B. Liu et al., "Effects of Prenatal PM 10 Exposure on Fetal Cardiovascular Malformations in Fuzhou, China: A Retrospective Case–Control Study," Environmental health perspectives, vol. 125, no. 5, pp. 057001-057001, 2017.
[6] J. D. Schwartz, Y. Wang, I. Kloog, M. a. Yitshak-Sade, F. Dominici, and A. Zanobetti, "Estimating the effects of PM 2.5 on life expectancy using causal modeling methods," Environmental health perspectives, vol. 126, no. 12, pp. 127002-127002, 2018.
[7] P. H. McMurry, M. F. Shepherd, and J. S. Vickery, Particulate matter science for policy makers: A NARSTO assessment. Cambridge University Press, 2004.
[8] "The Environmental Protection Administration's environmental open data platform." [Online]. Available: https://data.epa.gov.tw/.
[9] M. Han, R. Wei, and D. Li, "Multivariate chaotic time series analysis and prediction using improved nonlinear canonical correlation analysis," in 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), 2008, pp. 758–764-758–764.
[10] L. Pasanen and L. Holmström, "Scale space multiresolution correlation analysis for time series data," Computational Statistics, vol. 32, no. 1, pp. 197–218-197–218, 2017.
[11] M. Vyas, T. Guhr, and T. H. Seligman, "Multivariate analysis of short time series in terms of ensembles of correlation matrices," Scientific reports, vol. 8, no. 1, pp. 1–12-1–12, 2018.
[12] U. Kumar and V. K. Jain, "ARIMA forecasting of ambient air pollutants (O 3, NO, NO 2 and CO)," Stochastic Environmental Research and Risk Assessment, vol. 24, no. 5, pp. 751–760-751–760, 2010.
[13] J. K. Rekhi, P. Nagrath, R. Jain, and others, "Forecasting Air Quality of Delhi Using ARIMA Model," in Advances in Data Sciences, Security and Applications: Springer, 2020, pp. 315–325-315–325.
[14] B. Liu, Y. Jin, and C. Li, "Analysis and prediction of air quality in Nanjing from autumn 2018 to summer 2019 using PCR–SVR–ARMA combined model," Scientific Reports, vol. 11, no. 1, pp. 1–14-1–14, 2021.
[15] B.-C. Liu, A. Binaykia, P.-C. Chang, M. K. Tiwari, and C.-C. Tsao, "Urban air quality forecasting based on multi-dimensional collaborative Support Vector Regression (SVR): A case study of Beijing-Tianjin-Shijiazhuang," PloS one, vol. 12, no. 7, pp. e0179763-e0179763, 2017.
[16] H. Zhu and J. Hu, "Air quality forecasting using SVR with quasi-linear kernel," in 2019 International Conference on Computer, Information and Telecommunication Systems (CITS), 2019, pp. 1–5-1–5.
[17] N. H. Abd Rahman, M. H. Lee, M. T. Latif, and others, "Artificial neural networks and fuzzy time series forecasting: an application to air quality," Quality & Quantity, vol. 49, no. 6, pp. 2633–2647-2633–2647, 2015.
[18] J. C. Patni and H. K. Sharma, "Air quality prediction using artificial neural networks," in 2019 International Conference on Automation, Computational and Technology Management (ICACTM), 2019, pp. 568–572-568–572.
[19] H. Xie, F. Ma, and Q. Bai, "Prediction of indoor air quality using artificial neural networks," in 2009 fifth international conference on natural computation, 2009, vol. 2, pp. 414–418-414–418.
[20] A. Chakma, B. Vizena, T. Cao, J. Lin, and J. Zhang, "Image-based air quality analysis using deep convolutional neural network," in 2017 IEEE International Conference on Image Processing (ICIP), 2017, pp. 3949–3952-3949–3952.
[21] Q. Zhang, F. Fu, and R. Tian, "A deep learning and image-based model for air quality estimation," Science of the Total Environment, vol. 724, pp. 138178-138178, 2020.
[22] Z. Wu, Y. Wang, and L. Zhang, "MSSTN: Multi-Scale Spatial Temporal Network for Air Pollution Prediction," in 2019 IEEE International Conference on Big Data (Big Data), 2019, pp. 1547–1556-1547–1556.
[23] J. Xu, L. Chen, M. Lv, C. Zhan, S. Chen, and J. Chang, "HighAir: A Hierarchical Graph Neural Network-Based Air Quality Forecasting Method," arXiv preprint arXiv:2101.04264, 2021.
[24] P. Zhao and K. Zettsu, "MASTGN: Multi-Attention Spatio-Temporal Graph Networks for Air Pollution Prediction," in 2020 IEEE International Conference on Big Data (Big Data), 2020, pp. 1442–1448-1442–1448.
[25] C.-H. Hsu and F.-Y. Cheng, "Classification of weather patterns to study the influence of meteorological characteristics on PM2. 5 concentrations in Yunlin County, Taiwan," Atmospheric Environment, vol. 144, pp. 397–408-397–408, 2016.
[26] J.-H. Tsai, L.-P. Chang, and H.-L. Chiang, "Airborne pollutant characteristics in an urban, industrial and agricultural complex metroplex with high emission loading and ammonia concentration," Science of the total environment, vol. 494, pp. 74–83-74–83, 2014.
[27] P.-H. Tan, C. Chou, and C. C. K. Chou, "Impact of urbanization on the air pollution “holiday effect” in Taiwan," Atmospheric environment, vol. 70, pp. 361–375-361–375, 2013.
[28] J. Lee Rodgers and W. A. Nicewander, "Thirteen ways to look at the correlation coefficient," The American Statistician, vol. 42, no. 1, pp. 59–66-59–66, 1988.
[29] L. Li, J. Wu, N. Hudda, C. Sioutas, S. A. Fruin, and R. J. Delfino, "Modeling the concentrations of on-road air pollutants in southern California," Environmental science & technology, vol. 47, no. 16, pp. 9291–9299-9291–9299, 2013.
[30] Y.-C. Lin, S.-J. Lee, C.-S. Ouyang, and C.-H. Wu, "Air quality prediction by neuro-fuzzy modeling approach," Applied soft computing, vol. 86, pp. 105898-105898, 2020.
[31] J. Chung, C. Gulcehre, K. Cho, and Y. Bengio, "Empirical evaluation of gated recurrent neural networks on sequence modeling," arXiv preprint arXiv:1412.3555, 2014.
[32] M. Schuster and K. K. Paliwal, "Bidirectional recurrent neural networks," IEEE transactions on Signal Processing, vol. 45, no. 11, pp. 2673–2681-2673–2681, 1997.
[33] M.-T. Luong, H. Pham, and C. D. Manning, "Effective approaches to attention-based neural machine translation," arXiv preprint arXiv:1508.04025, 2015.
[34] S. Bai, J. Z. Kolter, and V. Koltun, "An empirical evaluation of generic convolutional and recurrent networks for sequence modeling," arXiv preprint arXiv:1803.01271, 2018.
[35] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778-770–778.
[36] S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman, "1D convolutional neural networks and applications: A survey," Mechanical Systems and Signal Processing, vol. 151, pp. 107398-107398, 2021.
[37] F. Yu and V. Koltun, "Multi-scale context aggregation by dilated convolutions," arXiv preprint arXiv:1511.07122, 2015.
[38] P. Ramachandran, B. Zoph, and Q. V. Le, "Searching for activation functions," arXiv preprint arXiv:1710.05941, 2017.
[39] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: a simple way to prevent neural networks from overfitting," The journal of machine learning research, vol. 15, no. 1, pp. 1929–1958-1929–1958, 2014.
[40] Ö. Yildirim, "A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification," Computers in biology and medicine, vol. 96, pp. 189–202-189–202, 2018.
[41] B. Kim, "Attention in Neural Networks - 1. Introduction to attention mechanism," 2020. [Online]. Available: https://buomsoo-kim.github.io/attention/2020/01/01/Attention-mechanism-1.md/.
[42] "UCI Machine Learning Repository." [Online]. Available: https://archive.ics.uci.edu/ml/datasets.php.
[43] Rmcab, "Bogot´a air quality monitoring network. website of environmental information." [Online]. Available: http://201.245.192.252:81/.
[44] W. Gornall and I. A. Strebulaev, "Squaring venture capital valuations with reality," Journal of Financial Economics, vol. 135, no. 1, pp. 120–143-120–143, 2020.
[45] "Keras：A deep learning library." [Online]. Available: https://keras.io/.
[46] "Tensorflow：A deep learning framework." [Online]. Available: https://www.tensorflow.org/.
[47] "Scikit-learn：A machine learning library in python." [Online]. Available: https://scikit-learn.org/.
[48] C. Cortes and V. Vapnik, "Support-vector networks," Machine learning, vol. 20, no. 3, pp. 273–297-273–297, 1995.
[49] C.-C. Chang and C.-J. Lin, "LIBSVM: a library for support vector machines," ACM transactions on intelligent systems and technology (TIST), vol. 2, no. 3, pp. 1–27-1–27, 2011.
[50] D. S. Broomhead and D. Lowe, "Radial basis functions, multi-variable functional interpolation and adaptive networks," 1988.
[51] S. Du, T. Li, Y. Yang, and S.-J. Horng, "Deep air quality forecasting using hybrid deep learning framework," IEEE Transactions on Knowledge and Data Engineering, 2019.
[52] G. Yang, H. Lee, and G. Lee, "A hybrid deep learning model to forecast particulate matter concentration levels in Seoul, South Korea," Atmosphere, vol. 11, no. 4, pp. 348-348, 2020.
[53] Q. Tao, F. Liu, Y. Li, and D. Sidorov, "Air pollution forecasting using a deep learning model based on 1D convnets and bidirectional GRU," IEEE access, vol. 7, pp. 76690–76698-76690–76698, 2019.
[54] C.-J. Huang and P.-H. Kuo, "A deep cnn-lstm model for particulate matter (PM2. 5) forecasting in smart cities," Sensors, vol. 18, no. 7, pp. 2220-2220, 2018.
[55] F. Franceschi, M. Cobo, and M. Figueredo, "Discovering relationships and forecasting PM10 and PM2. 5 concentrations in Bogotá, Colombia, using artificial neural networks, principal component analysis, and k-means clustering," Atmospheric Pollution Research, vol. 9, no. 5, pp. 912–922-912–922, 2018.
[56] P. J. Huber, "Robust estimation of a location parameter," in Breakthroughs in statistics: Springer, 1992, pp. 492–518-492–518.
[57] F. Chollet and J. J. Allaire, "RStudio AI Blog: Time Series Forecasting with Recurrent Neural Networks," ed, 2017.
[58] M. Bramer, "Data for data mining," in Principles of data mining: Springer, 2016, pp. 223-223.