Journal of the Korea Academia-Industrial cooperation Society (한국산학기술학회논문지)

The Korea Academia-Industrial cooperation Society (한국산학기술학회)
권호 표지
DOI QR코드

DOI QR Code

Real-time PM10 Concentration Prediction LSTM Model based on IoT Streaming Sensor data

IoT 스트리밍 센서 데이터에 기반한 실시간 PM10 농도 예측 LSTM 모델

  • Kim, Sam-Keun (Computer System Institute, Department of Computer Engineering, Hankyong National University) ;
  • Oh, Tack-Il (Department of Computer Engineering, Hankyong National University)
  • 김삼근 (한경대학교 컴퓨터시스템연구소&컴퓨터공학과) ;
  • 오택일 (한경대학교 컴퓨터공학과)
  • Received : 2018.09.28
  • Accepted : 2018.11.02
  • Published : 2018.11.30
Download PDF
⟨ Previous Next ⟩

Abstract

Recently, the importance of big data analysis is increasing as a large amount of data is generated by various devices connected to the Internet with the advent of Internet of Things (IoT). Especially, it is necessary to analyze various large-scale IoT streaming sensor data generated in real time and provide various services through new meaningful prediction. This paper proposes a real-time indoor PM10 concentration prediction LSTM model based on streaming data generated from IoT sensor using AWS. We also construct a real-time indoor PM10 concentration prediction service based on the proposed model. Data used in the paper is streaming data collected from the PM10 IoT sensor for 24 hours. This time series data is converted into sequence data consisting of 30 consecutive values from time series data for use as input data of LSTM. The LSTM model is learned through a sliding window process of moving to the immediately adjacent dataset. In order to improve the performance of the model, incremental learning method is applied to the streaming data collected every 24 hours. The linear regression and recurrent neural networks (RNN) models are compared to evaluate the performance of LSTM model. Experimental results show that the proposed LSTM prediction model has 700% improvement over linear regression and 140% improvement over RNN model for its performance level.

최근 사물인터넷(IoT)의 등장으로 인터넷에 연결된 다양한 기기들에 의해 대규모의 데이터가 생성됨에 따라 빅데이터 분석의 중요성이 증가하고 있다. 특히 실시간으로 생성되는 대규모의 IoT 스트리밍 센서 데이터를 분석하여 새로운 의미 있는 미래 예측을 통해 다양한 서비스를 제공하는 것이 필요하게 되었다. 본 논문은 AWS를 활용하여 IoT 센서로부터 생성되는 스트리밍 데이터에 기반하여 실시간 실내 PM10 농도 예측 LSTM 모델을 제안한다. 또한 제안 모델에 따른 실시간 실내 PM10 농도 예측 서비스를 구축한다. 논문에 사용된 데이터는 PM10 IoT 센서로부터 24시간 동안 수집된 스트리밍 데이터이다. 이를 LSTM의 입력 데이터로 사용하기 위해 PM10 시계열 데이터로부터 30개의 연속된 값으로 이루어진 시퀀스 데이터로 변환한다. LSTM 모델은 바로 인접한 공간으로 이동해 가는 슬라이딩 윈도우 프로세스를 통하여 학습한다. 또한 모델의 성능 개선을 위해 24시간마다 수집한 스트리밍 데이터에 대해 점진적 학습 방법을 적용한다. 제안한 LSTM 모델의 성능을 평가하기 위해 선형회귀 모델 및 순환형 신경망(RNN) 모델과 비교한다. 실험 결과는 제안한 LSTM 예측 모델이 선형 회귀보다 700%, RNN 모델보다는 140% 성능 개선이 있음을 보여주었다.

Keywords

SHGSCZ_2018_v19n11_310_f0001.png 이미지

Fig. 1. RNN Architecture [9]

SHGSCZ_2018_v19n11_310_f0002.png 이미지

Fig. 2. LSTM Architecture [2]

SHGSCZ_2018_v19n11_310_f0003.png 이미지

Fig. 3. Training data acquisition from AWS IoT

SHGSCZ_2018_v19n11_310_f0004.png 이미지

Fig. 4. PM10 raw data

SHGSCZ_2018_v19n11_310_f0005.png 이미지

Fig. 5. Flow chart of the proposed PM10 prediction LSTM model

SHGSCZ_2018_v19n11_310_f0006.png 이미지

Fig. 6. Flow chart of real-time PM10 prediction service

SHGSCZ_2018_v19n11_310_f0007.png 이미지

Fig. 7. LSTM loss

SHGSCZ_2018_v19n11_310_f0008.png 이미지

Fig. 8. Example of PM10 actual/prediction graphs

Table 1. LSTM_with_peepholes Parameter Settings: sliding_ window=30, window_shift=1

SHGSCZ_2018_v19n11_310_t0001.png 이미지

Table 2. Model Evaluation

SHGSCZ_2018_v19n11_310_t0002.png 이미지

References

  1. Amazon Web Service, https://docs.aws.amazon.com/
  2. Christopher Olah, Understanding LSTM Networks, 2015, Available From: https://www.cse.iitk.ac.in/users/sigml/lec/Slides/LSTM.pdf.
  3. Korea Environment Corporation(KECO), Available From: http://www.airkorea.or.kr/dictionary_3. (accessed Sept., 12, 2018)
  4. M. S. Souza, P. Coelho, A. da Silva, A. Pozza, "Using Ensembles of Artificial Neural Networks to Improve PM10 Forecasts", Chemical Engineering Transactions, Vol. 43, pp. 2161-2166, 2015. DOI: https://doi.org/10.3303/CET1543361
  5. B. Oancea, S. C. Ciucu, "Time series forecasting using neural networks", Proceedings of the CKS 2013 International Conference, pp. 1402-1408, 2014, Available From: https://arxiv.org/ftp/arxiv/papers/1401/1401.1333.pdf.
  6. Yanchen Liua, Zhe Wangc, Zhongchen Zhang, Jiajie Hong, Borong Lin, "Investigation on the Indoor Environment Quality of health care facilities in China", Building and Environment 141, pp. 273-287, 2018, DOI: https://doi.org/10.1016/j.buildenv.2018.05.054
  7. Gyu-Sik Kim, Youn-Suk Son, Jai-Hyo Lee, In-Won Kim, Jo-Chun Kim, Joon-Tae Oh, Hiesik Kim, "Air Pollution Monitoring and Control System for Subway Stations Using Environmental Sensors", Hindawi Publishing Corporation Journal of Sensors, Vol. 2016. DOI: https://doi.org/10.1155/2016/1865614
  8. Jin-Ho Noh, Han-Ho Tack, "The Implementation of the Fine Dust Measuring System based on Internet of Things(IoT)", Journal of the Korea Institute of Information and Communication Engineering, Vol. 21, No. 4, pp. 829-835, 2017, DOI: http://www.dbpia.co.kr/Article/NODE07158660 https://doi.org/10.6109/JKIICE.2017.21.4.829
  9. Recurrent neural network, Wikipedia, 2018, https://en.wikipedia.org/wiki/Recurrent_neural_network
  10. Xavier Glorot, Yoshua Bengio, "Understanding the difficulty of training deep feedforward neural networks", Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, PMLR 9, pp. 249-256, 2010, Available From: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.207.2059&rep=rep1&type=pdf
  11. Hyperbolic function, Wikipedia, 2018, https://en.wikipedia.org/wiki/Hyperbolic_function
  12. AdamOptimizer, https://www.tensorflow.org/api_docs/python/tf/train/AdamOptimizer
  13. Root-mean-square deviation, Wikipedia, 2018, Available From: https://en.wikipedia.org/wiki/Root-mean-square_deviation.
  14. D. Dunea, S. Iordache, C. Ianache, "Relationship between airborne particulate matter and weather conditions in targoviste urban area during cold months", Roumanian Journal of Chemistry, Vol. 60, pp. 595-601, 2015, Available From: https://www.researchgate.net/publication/284735048
  15. WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide, 2005, Available From: http://apps.who.int/iris/bitstream/handle/10665/69477/WHO_SDE_PHE_OEH_06.02_eng.pd
  16. NAAQS Table, United States Environmental Protection Agency, 2016, Available From: https://www.epa.gov/criteria-air-pollutants/naaqs-table
  17. D. Vlachogiannis, A. Sfetsos, "Time series forecasting of hourly PM10 values: model intercomparison and the development of localized linear approaches", WIT Transactions on Ecology and the Environment, Vol. 86, pp. 85-94, 2006 DOI: https://doi.org/10.2495/AIR06009
  18. Hazrul Abdul Hamid, Ahmad Shukri Yahaya, Nor Azam Ramli, Ahmad Zia Ul-Saufie and Mohd Norazam Yasin, "Short Term Prediction of PM10 Concentrations Using Seasonal Time Series Analysis", MATEC Web of Conferences, Vol. 47, 2016. DOI: https://doi.org/10.1051/matecconf/20164705001
  19. H. Weizhen, L. Zhengqiang, Z. Yuhuan, X. Hua, Z. Ying, L. Kaitao, L. Donghui, W. Peng, M. Yan, "Using support vector regression to predict PM10 and PM2.5", In IOP Conference Series: Earth and Environmental Science, Vol. 17, 2014. DOI: https://doi.org/10.1088/1755-1315/17/1/012268
  20. J. Chung, C. Gulcehre, K. H. Cho, Y. Bengio, "Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling", Neural Information Processing Systems, 2014, Available From: https://arxiv.org/pdf/1412.3555.pdf
  21. Incremental learning, Wikipedia, 2018, https://en.wikipedia.org/wiki/Incremental_learning/
  22. Docker, https://www.docker.com/
  23. TensorFlow Serving, https://www.tensorflow.org/serving/
  24. TensorFlow, https://www.tensorflow.org/