PM2.5的影響因子很多，減少PM2.5的排放量是國際重視的議題之一。然而，近年來的研究指出衍生性PM2.5來源之一是由氨氣和工廠排放的廢氣經過複雜的化學反應而生成，因此，本研究收集政府提供的相關開放資料，包括空氣監測站的氣象資料、登記的工業排放的空氣汙染資料、畜牧業資料，想探討的議題有兩項，第一項議題是使用Apache Spark當作雲端運算平台，研究調查空氣監測站附近的列管空汙污染源資料和畜牧業動物數量，對於PM2.5年平均的影響。第二項議題是研究調查環保署列管空汙污染源開放資料和農委會畜牧業開放資料，使用Tensorflow進行深度學習模型研究，探討相關衍生性因子對PM2.5濃度是否有造成影響。實驗結果顯示空氣監測站附近的工業排放空汙資料和畜牧業動物數量資料，對於PM2.5年平均的Spearman's correlation coefficient都介於0到1之間，代表正相關。深度學習實驗顯示工業資料加上畜牧業資料加上氣象資料的深度學習模型PM2.5濃度的MSE分類準確度為0.75，對比只有氣象資料的深度學習模型PM2.5濃度MSE分類準確度為1.5。因此得出結論，氣象因子和工業因子和畜牧業因子對於研究區域的PM2.5濃度的變化是可能的影響項目之一，希望提供政府單位相關的決策及可以藉由在工廠和牧場裝設相關空氣監測器，分析空氣的品質資料，為改善環境之相關依據，減少PM2.5的排放量，降低民眾罹患心血管疾病的機率。

There are many factors that influence PM2.5, reducing the emission of PM2.5 is one of the subjects of the world's interest. However, it is indicated recently that one of the sources of the secondary PM2.5 is the complex chemical reaction of NH3 and the exhaust gases emitted from factories. Therefore, this study collects the open data provided by the government, including the weather data of air monitoring stations, the air pollution data of registered industrial discharge and stock raising data. There are two subjects to be discussed. Subject 1 is to use Apache Spark as Cloud computing platform to study the effect of controlled air pollution source data and size of animal nearby the air monitoring station on the annual mean PM2.5. Subject 2 is to study the open data of controlled air pollution sources from the Environmental Protection Administration and the open data of stock raising from the Council of Agriculture, the Tensorflow is used to study deep learning model to discuss whether the related derived factors affect the PM2.5 concentration or not. The experimental results show that the Spearman's correlation coefficient of the air pollution data of industrial discharge and the size of animal nearby the air monitoring station for the annual mean PM2.5 is 0 to 1, representing positive correlation. The deep learning experiment shows that the MSE classification accuracy of PM2.5 concentration of the deep learning model with industrial data + stock raising data + weather data is 0.75, whereas the MSE classification accuracy of PM2.5 concentration of the deep learning model only with weather data is 1.5. Therefore, the meteorological factor, industrial factor and stock raising factor may influence the PM2.5 concentration in the study area, hoping to provide references for the government bodies to make decisions and to install related air monitors in factories and livestock farms to analyze the air quality data, so as to improve the environment, reduce the emission of PM2.5 and reduce the probability of suffering from cardiovascular disease.

[1] M. Dong, D. Yang, Y. Kuang, D, He, S. Erdald, D. Kenskie, “PM2.5 concentration prediction using hidden semi-Markov model-based times series data mining”, Expert Systems with Applications, vol. 36, pp. 9046-9055, 2009
[2] P. Perez, E. Gramsch,” Forecasting hourly PM2.5 in Santiago de Chile with emphasis on night episodes”, Atmospheric Environment, vol. 124, pp. 22-27, 2016
[3] D. J. Lary, F. S. Faruque, N. Malakar, A. Moore, B. Roscoe, Z. L. Adams, Y. Eggelston, “Estimating the global abundance of ground level presence of particulate matter (PM2.5),” Geospatial Health, vol. 8(3), pp. 611-630, 2014
[4] N. Zhang, H. Chen, X. Chen and J. Chen, “Semantic Framework of Internet of Things for Smart Cities: Case Studies,” Sensors, vol. 16, 2016
[5] X. Zhang, X. Rui, X. Xia, X. Bai, W. Yin, J. Dong, ”A hybrid model for short-term air pollutant concentration forecasting”, Service Operations And Logistics, And Informatics (SOLI), 2015
[6] B. T. Ong, K. Sugiura , K. Zettsu, ” Dynamic pre-training of Deep Recurrent Neural Networks for predicting environmental monitoring data”, IEEE International Conference on Big Data , 2014
[7] H. Zhu, X. Lu, “The Prediction of PM2.5 Value Based on ARMA and Improved BP Neural Network Model”, Intelligent Networking and Collaborative Systems (INCoS), 2016
[8] B. Zou, M. Wang, N. Wan, J. G. Wilson, X. Fang, Y. Tang “Spatial modeling of PM2.5 concentrations with a multifactoral radial basis function neural network," Environmental Science and Pollution Research, vol. 22(14), pp.10395-10404.2015
[9] B .Chen, X. Wang, 1. Yu, H. Wang, Y. Li, J. Chen, J. Zhu, H. N. and L. Hou, “Prediction of PM2.5 Concentration in a Agricultural Park Based on BP Artificial Neural Network,” Advance Journal of Food Science and Technology, vol.11(4), pp.274-280, 2016
[10] S. E. Bauer, K. Tsigaridis, R. Miller, ”Significant atmospheric aerosol pollution caused by world food cultivation,” Geophysical Research Letters, vol.43, pp. 5394–5400, 2016
[11] C. L. Heald, J. L. Collett Jr., T. Lee, K. B. Benedict, F. M. Schwandner, Y. Li, L. Clarisse, D. R. Hurtmans, M. Van Damme, C. Clerbaux, P.-F. Coheur, S. Philip, R. V. Martin, and H. O. T. Pye,” Atmospheric ammonia and particulate inorganic nitrogen over the United States,” Atmos. Chem. Phys, pp.10295–10312, 2012
[12] L. Li, M. Kumar, C. Zhu, J. Zhong, Joseph S. Francisco, and Xiao Cheng Zeng,” Near-Barrierless Ammonium Bisulfate Formation via a Loop- Structure Promoted Proton-Transfer Mechanism on the Surface of Water”, Journal of the American chemical society, pp. 1816−1819, 2016
[13] C. Zhang, J. Yan, C. Li, X. Rui, L. Liu, R. Bie, “On Estimating Air Pollution from Photos Using Convolutional Neural Network,” ACM, pp.297-301
[14] B. T. Ong, K. Sugiura, K. Zettsu, “Dynamically pre-trained deep recurrent neural networks using environmental monitoring data for predicting PM2.5,” Neural Computing and Applications, vol.27, pp 1553–1566, 2016,
[15] R.A. Bahari , R. A. Abbaspour, P. Pahlavani, “PREDICTION OF PM2.5 CONCENTRATIONS USING TEMPERATURE INVERSION EFFECTS BASED ON AN ARTIFICIAL NEURAL NETWORK,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XL-2/W3, 2014
[16] X. Li, L. Yuan, H. Shao, T. Chi, “Deep learning architecture for air quality predictions”, Environmental Science and Pollution Research, vol.23, pp. 22408–22417, 2016
[17] N. Zhang, H. Chen, X. Chen and J. Chen, “Semantic Framework of Internet of Things for Smart Cities: Case Studies,” Sensors, vol. 16, 2016
[18] F Rosenblatt, “The perceptron: A probabilistic model for information storage and organization in the brain”, Psychological Review, Vol. 65, pp. 386-408, 1958
[19] I. Ovidiu, “Applications of Support Vector Machines in Chemistry”, in Reviews in Computational Chemistry, Volume 23, 2007.
[20] D. S. Broomhead, D. Lowe,"Multivariable functional interpolation and adaptive networks". Complex Systems, pp.321–355, 1998
[21] G. E. Hinton, R. R. Salakhutdinov, "Reducing the Dimensionality of Data with Neural Networks", Science, vol. 313, pp.504–507, 2006
[22] M. Pharr, R. Fernando, “Gpu gems 2: programming techniques for high-performance graphics and general-purpose computation”, 2005
[23] Seide, Frank, G. Li, D. Yu, "Conversational Speech Transcription Using Context-Dependent Deep Neural Networks", Interspeech, 2011.
[24] B. Huval, T. Wang, S. Tandon, “An Empirical Evaluation of Deep Learning on Highway Driving”, Computer Science, 2015
[25] C.Michael, “AlphaGo and Monte Carlo tree search: the simulation optimization perspective”, Proceedings of the 2016 Winter Simulation Conference, 2016
[26] Zo. Tuske, M. Ali Tahir, R. Schlüter “H. Ney Integrating Gaussian mixtures into deep neural networks: Softmax layer with hidden variables”, Speech and Signal Processing, 2015
[27] F Harrell ,”Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis”, 2015
[28] D. P. Kingma, Kingma, J. L. Ba,”Adam: A Method for Stochastic Optimization”, ICLR,2015
[29] R. Hahnloser, R. Sarpeshkar, M. A. Mahowald, R. J. Douglas, H.S. Seung, “Digital selection and analogue amplification coesist in a cortex-inspired silicon circuit. Nature”, vol. 405. pp. 947–951, 2000.
[30] Tieleman, Tijmen and Hinton, Geoffrey (2012). Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude. COURSERA: Neural Networks for Machine Learning
[31] J. Schmidhuber, “Deep learning in neural networks: An overview”, Neural Networks, vol. 61, pp.85-117, 2015
[32] W. Hou, X. Gao, D. Tao, Xuelong Li, “Blind Image Quality Assessment via Deep Learning”, IEEE Transactions on Neural Networks and Learning Systems, vol.26, pp.1275-1286, 2015
[33] B. Gaonkar, D. Hovda, N. Martin, L. Macyszyn, ” Deep learning in the small sample size setting: Cascaded feed forward neural networks for medical image segmentation”, SPIE Proceedings, vol.9785, 2016
[34] H. W. Ng, V. D. Nguyen, V. Vonikakis, S. Winkler, “Deep Learning for Emotion Recognition on Small Datasets Using Transfer Learning”, ACM. 2015
[35] K. M. Kumar, A. R. M. Reddy, “A fast DBSCAN clustering algorithm by accelerating neighbor searching using Groups method,” Pattern Recognition, vol.58, pp.39–48, 2016
[36] P. Viswanatha, V. S. Babub, “Rough-DBSCAN: A fast hybrid density based clustering method for large data sets”, Pattern Recognition Letters, vol. 30, pp.1477-1488, 2009
[37] M. T. Puth, M. Neuhausera, G. D. Ruxton, “Effective use of Spearman's and Kendall's correlation coefficients for association between two measured traits,” Animal Behaviour, vol. 102, pp.77-84, 2015
[38] Y. W. Laih,”European Journal of Operational Research, vol. 232, pp. 375-382, 2014