MPC-TimesNet: 다변량 주기성 정보를 반영한 대기오염도 장기예측 모델
MPC-TimesNet: Time-Series Long-Term Forecasting Air Pollution with Multiple Periodicity Consideration Model
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
도심 지역에서 발생하는 대기오염은 환경, 국민건강, 사회경제, 정책 결정 등에 중대한 영향을 끼친다. 전 세계적으로 지속적인 대기질 정보의 모니터링과 예측 연구의 필요성이 대두됨에 따라 최근 대기오염도 예측에서 Transformer에 기반한 방법들이 연구되고 있다. 하지만 이 방법들은 각각의 주기에 있는 local 정보를 고려함에 한계점이 있다. 이를 해결하기 위해서, 본 연구에서는 높은 주기성 정보와 낮은 주기성 정보를 활용하여 local 정보를 보다 면밀히 반영할 수 있는 대기오염도 예측을 위한 MPC(Multiple Periodicity Consideration)-TimesNet 모델을 제안한다. 제안 모델은 기존 TimesNet 모델에 HFEM(High Frequencies Extraction Module)을 추가함으로써 높은 주기 정보를 세밀하게 고려할 수 있도록 하고, 정보량이 낮아질수록 학습에 반영되는 표현력을 줄여나감으로써 노이즈가 생기는 문제를 완화하고자 한다. 제안 모델을 실험적으로 검증하기 위해 서울특별시의 세 지역구(송파구, 영등포구, 중구)에서 5종 대기오염원(이산화질소, 오존, 이산화탄소, 아황산가스, 미세먼지) 데이터를 사용하였고, 실험 결과, 제안 모델의 평균제곱오차는 0.3615, 0.3277, 0.3893, 평균절대오차는 0.3764, 0.3864, 0.39를 기록하며 모든 지역구에서 SOTA를 달성하였다. 베이스라인 모델인 TimesNet보다 성능이 높아진 것을 통해 HFEM 모듈이 대기오염도 예측에 효과적임을 확인했다.
Abstract
Urban air pollution has a significant impact on environment, public health, social economic, and policy decision. In accordance with increasing global necessity of constant monitoring and accurate forecasting of air pollution among governments and institutions, research into applying transformer-based models to predict pollutants preceded. However, the models have been limited in their ability to thoroughly consider local information within periodicity, which is one of the key inherent patterns in pollution data. To address the limitation, this paper proposes MPC(Multiple Periodicity Consideration)-TimesNet model designed to handle intricate locality information within periodicity by considering not only low frequencies information, but also that of high frequencies. The proposed model added HFEM(High Frequency Extraction Module) to the previous TimesNet model in order to exquisitely account for high frequencies data. In addition, we lowered the size of representations as the self-information gets smaller to address noise issues. To evaluate our model, we collected data of five air pollutants (NO2, OZ, CO2, SO2, PM10) among three regions―Songpa-gu, Youngdeungpo-gu, Jung-gu―in Seoul. As a result of the experiment, the proposed model recorded 0.3615, 0.3277, 0.3893 in mean square error and 0.3764, 0.3864, 0.39 in mean absolute error, respectively. Our model outperforms twelve previous SOTA models including baseline in all regions. In conclusion, we verified that summating HFEM increased the effectiveness of the model in forecasting air pollution by comparing their performance to the baseline, TimesNet.
참고문헌 (41)
- K. Naddafi, "Health impact assessment of air pollution in megacity of Tehran, Iran", Iranian Journal of environmental Health Science & Engineering, 9, pp.1~7 (2012)
- L. Zhang, "Transpacific transport of ozone pollution and the effect of recent Asian emission increases on air quality in North America: an integrated analysis using satellite, aircraft, ozonesonde, and surface observation", Atmospheric Chemistry and Physics, 8, pp.6117~6136 (2008)
- Ch. Vlachokostas, "Economic damages of ozone air pollution to crops using combined air quality and GIS modelling", Atmospheric Environment, 44, pp.3352~3361 (2010)
- A. Maghrabi, "Impact of the March 2009 dust event in Saudi Arabia on aerosol optical properties, meteorological parameters, sky temperature and emissivity", Atmospheric Environment, 45, pp.2164~2173 (2011)
- P. Hou, "Long-term changes in extreme air pollution meteorology and the implications for air quality", Scientific Reports, 6, pp.23792~ (2016)
- F. Biancofiore, "Recursive neural network model for analysis and forecast of PM10 and PM2. 5", Atmosphere Pollution Research, 10, pp.8~15 (2017)
- S. M. S. Cabaneros, "Hybrid artificial neural network models for effective prediction and mitigation of urban roadside NO2 pollution", Energy Procedia, 142, pp.3524~3530 (2017)
- A. Coman, "Hourly ozone prediction for a 24-h horizon using neural networks", Environmental Modelling Software, 23, pp.1407~1421 (2008)
- S. D. Lightstone, "Comparing CMAQ forecasts with a neural network forecast model for PM2.5 in New York", Atmosphere, 8(9), pp.161~ (2017)
- A. Rahimi, "Short-term prediction of NO2and NO x concentrations using multilayer perceptron neural network: a case study of Tabriz", Iran, Ecological Processes, 6, pp.1~9 (2017)
- B. S. Freeman, "Forecasting air quality time series using deep learning", Journal of the Air & Waste Management Association, 68, pp.866~886 (2018)
- R. Espinosa, "A time series forecasting based multi-criteria methodology for air quality prediction", Applied Soft Computing, 113, pp.107850~ (2021)
- S. H. Hickman, "Forecasting European ozone air pollution with Transformers", Tackling Climate Change with Machine Learning: workshop at NeurIPS (2022)
- C. Fan, "Multi-horizon time series forecasting with temporal attention learning", Proceedings of the 25th Association for Computing Machinary Special Interest Group on Knowledge Discovery in Data International conference on knowledge discovery & data mining, pp.2527~2535 (2019)
- Y. Liang, "Airformer : Predicting nationwide air quality in china with transformers", Proceedings of the AAAI Conference on Artificial Intelligence, 37(12), pp.14329~14337 (2023)
- S. Li, "Enhancing the locality and breaking the memory bottleneck of transformer on time series forecasting", Advances in neural information processing systems, 32 (2019)
- H. Wu, "Timesnet: Temporal 2d-variation modeling for general time series analysis", arXiv preprint arXiv:2210.02186, ~ (2023)
- S. G. Gocheva-Ilieva, "Time series analysis and forecasting for air pollution in small urban area : an SARIMA and factor analysis approach", Stochastic environmental research and risk assessment, 28, pp.1045~1060 (2014)
- H. Maleki, "Air pollution prediction by using an artificial neural network model", Clean technologies and environmental policy, 21, pp.1341~1352 (2019)
- Y. S. Chang, "An LSTM-based aggregated model for air pollution forecasting", Atmospheric Pollution Research, 11, pp.1451~1463 (2020)
- C. Wen, "A novel spatiotemporal convolutional long short-term neural network for air pollution prediction", Science of the total environment, 654, pp.1091~1099 (2019)
- H. Feng, "A novel encoder-decoder model based on Autoformer for air quality index prediction", PLoS ONE, 18, pp.e0284293~ (2023)
- K. Cai, "Improving air pollutant prediction in Henan Province, China, by enhancing the concentration prediction accuracy using autocorrelation errors and an Informer deep learning model", Sustainable Environment Research, 33, pp.1~13 (2023)
- Y. Feng, "Spatiotemporal informer: A new approach based on spatiotemporal embedding and attention for air quality forecasting", Environmental Pollution, 336, pp.122402~ (2023)
- 권혁재, "DTD Transformer-MAF: 인접 시점의 다양한 맥락 정보를 강조한 우리나라 주요 수출 8개국의 장기 원화 환율 분포 추정", 한국지식정보기술학회 논문지, 18(4), pp.907~921 (2023), 한국지식정보기술학회[DOI]
- 김영진, "장기 변동성을 반영하기 위한 확산 모델을 활용한 주가 지수 확률 분포 예측", 한국지식정보기술학회 논문지, 18(1), pp.141~155 (2023), 한국지식정보기술학회[DOI]
- J. W. Cooley, "The fast Fourier transform and its applications", IEEE Transactions on Education, 12, pp.27~34 (1969)
- C. Szegedy, "Going deeper with convolutions", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.1~9 (2015)
- C. Szegedy, "Rethinking the inception architecture for computer vision", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.2818~2826 (2016)
- K. He, "Deep residual learning for image recognition", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.770~778 (2016)
- A. Vaswani, "Attention is all you need", Advances in neural information processing systems, 30 (2017)
- H. Wu, "Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting", Advances in Neural Information Processing Systems, 34, pp.22419~22430 (2021)
- Y. Zhang, "Crossformer:transformer utilizing cross-dimension dependency for multivariate time series forecasting", The Eleventh International Conference on Learning Representations (2022)
- T. Zhou, "Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting", International Conference on Machine Learning, pp.27268~27286 (2022)
- Y. Liu, "Non-stationary transformers: Rethinking the stationarity in time series forecasting", arXiv preprint arXiv:2205.14415, ~ (2022)
- Y. Nie, "A time series is worth 64words: Long-term forecasting with transformers", arXiv preprint arXiv:2211.14730, ~ (2022)
- H. Zhou, "Informer:Beyond efficient transformer for long sequence time-series forecasting", Proceedings of the AAAI conference on artificial intelligence, 35(12), pp.11106~11115 (2021)
- N. Kitaev, "Reformer: The efficient transformer", arXiv preprint arXiv:2001.04451, ~ (2020)
- T. Zhang, "Less is more: Fast multivariate time series forecasting with light sampling-oriented mlp structures", arXiv preprint arXiv:2207.01186, ~ (2022)
- A. Zeng, "Are transformers effective for time series forecasting?", Proceedings of the AAAI conference on artificial intelligence, 37(9), pp.11121~11128 (2023)
- T. Zhou, "Film: Frequency improved legendre memory model for long-term time series forecasting", Advances in Neural Information Processing Systems, 35, pp.12677~12690 (2022)