[논문]앙상블 기반 모델을 이용한 서울시 PM2.5 농도 예측 및 분석

류민지; 손상훈; 김진수

doi:10.7780/kjrs.2022.38.6.1.16

[국내논문] 앙상블 기반 모델을 이용한 서울시 PM2.5 농도 예측 및 분석
Prediction and Analysis of PM2.5 Concentration in Seoul Using Ensemble-based Model 원문보기

대한원격탐사학회지 = Korean journal of remote sensing, v.38 no.6 pt.1, 2022년, pp.1191 - 1205

류민지 (부경대학교 지구환경시스템과학부 공간정보시스템공학전공) , 손상훈 (부경대학교 지구환경시스템과학부 공간정보시스템공학전공) , 김진수 (부경대학교 지구환경시스템과학부 공간정보시스템공학전공)

초록
AI-Helper

복잡하고 광범위한 원인을 가진 대기오염물질 중 particulate matter (PM)은 입자의 크기에 따라 분류된다. 그 중 PM_2.5는 그 크기가 매우 작아 사람이 흡입하면 인간의 호흡기나 심혈관에 질병을 유발할 수 있다. 이러한 위험에 대비하기 위해서는 국가 중심의 관리와 사전에 예방할 수 있는 모니터링 및 예측이 중요하다. 본 연구는 고농도 미세먼지의 발생이 잦은 서울시의 PM_2.5를 local data assimilation and prediction system (LDAPS) 기상 관련 인자 15가지와 aerosol optical depth (AOD), 화학인자 4가지를 독립변수로 하여 앙상블 모델 두 가지 random forest (RF)와 extreme gradient boosting (XGB)로 예측하고자 하였다. 예측에 사용된 두 모델의 성능 평가와 인자 중요도 평가를 수행하였으며, 계절별 모델 분석도 수행하였다. 예측 정확도 결과, RF가 R² = 0.85, XGB가 R² = 0.91의 높은 예측 정확도를 보이며 XGB가 RF보다 PM_2.5 예측에 적합한 모델임을 확인하였다. 계절별 모델 분석 결과, 봄에 농도가 높은 관측 값과 비교하여 예측 수행이 잘 되었다고 할 수 있다. 본 연구는 다양한 인자를 이용하여 서울시의 PM_2.5를 예측하였고, 좋은 성능을 보이는 앙상블 기반의 PM_2.5 예측 모델을 구축하였다.

Abstract ▼ AI-Helper

Particulate matter(PM) among air pollutants with complex and widespread causes is classified according to particle size. Among them, PM2.5 is very small in size and can cause diseases in the human respiratory tract or cardiovascular system if inhaled by humans. In order to prepare for these risks, state-centered management and preventable monitoring and forecasting are important. This study tried to predict PM2.5 in Seoul, where high concentrations of fine dust occur frequently, using two ensemble models, random forest (RF) and extreme gradient boosting (XGB) using 15 local data assimilation and prediction system (LDAPS) weather-related factors, aerosol optical depth (AOD) and 4 chemical factors as independent variables. Performance evaluation and factor importance evaluation of the two models used for prediction were performed, and seasonal model analysis was also performed. As a result of prediction accuracy, RF showed high prediction accuracy of R2 = 0.85 and XGB R2 = 0.91, and it was confirmed that XGB was a more suitable model for PM2.5 prediction than RF. As a result of the seasonal model analysis, it can be said that the prediction performance was good compared to the observed values with high concentrations in spring. In this study, PM2.5 of Seoul was predicted using various factors, and an ensemble-based PM2.5 prediction model showing good performance was constructed.

주제어

표/그림 (10)

그림 Fig. 1. Study area and 40 AQMS in Seoul.
표 Table 1. Data used for predictors of PM_2.5
그림 Fig. 2. Raster data for AOD and LDAPS (26 February 2018).
그림 Fig. 3. Scatterplot of model performance: (a) RF train, (b) RF test, (c) XGB train, and (d) XGB test.
표 Table 2. A summary of model performance results
그림 Fig. 4. Feature importance (a) RF and (b) XGB.
그림 Fig. 5. Comparison of observed and predicted values: (a) comparison of spring observations and model predicts, (b) comparison of summer observations and model predicts, (c) comparison of fall observations and model predicts, and (d) comparison of winter observations and model predicts. Values are given as means.
그림 Fig. 6. RF prediction accuracy as a seasonal scatterplot: (a) Spring RF model performance Scatterplot,(b) Summer RF model performance scatterplot, (c) Fall RF model performance scatterplot, and(d) Winter RF model performance scatterplot.
그림 Fig. 7. XGB prediction accuracy as a seasonal scatterplot: (a) Spring XGB model performance scatterplot, (b) Summer XGB model performance scatterplot, (c) Fall XGB model performance scatterplot, and (d) Winter XGB model performance scatterplot.
표 Table 4. Seasonal analysis result

AI 본문요약
AI-Helper

대상 데이터

두 데이터는 대한민국이 위치한 픽셀의 해당 값을 AQMS 위치 데이터를 기반으로 지상 관측 인자에 join 하여 최종 데이터 셋을 구축하였다. 모든 입력 데이터를 AQMS 지점에 매칭한 후 25개의 인자 중 결측 값이 하나라도 포함되는 날의 데이터는 모두 제거하였고 최종 데이터는 총 15,300개로 구축되었다.
데이터셋은 XGB 모델에서 train_test_split 기능으로 분할하여 RF와 XGB에 동일한 train, test 데이터셋을 적용하였다. 최종적으로 train 12,240개 test 3,060개로 분할된 데이터셋을 이용하여 모델의 훈련과 검증을 수행하였다.

데이터처리

Grid-Search CV 기법은 사용자가 직접 모델의 하이퍼 파라미터의 값을 리스트로 입력하면 값에 대한 경우의 수마다 예측 성능을 평가하고 비교하며 최적의 파라미터 값을 찾는 과정을 진행하는 방식이다. RF에 5-fold CV와 Grid-search CV를 적용해 RF의 n_estimators에 대한 파라미터 최적화 과정을 거쳐 최종 RF 모델 구축을 수행하였다. XGB 또한 K-fold CV와 Grid-search CV를 적용하였다.
구축된 모델의 훈련과 검증을 위하여 데이터를 train과 test로 분할하였다. 데이터셋은 XGB 모델에서 train_test_split 기능으로 분할하여 RF와 XGB에 동일한 train, test 데이터셋을 적용하였다.

이론/모형

RF 모델의 과적합 방지와 모델의 정확도 향상을 위한 파라미터 최적화 과정에서 k-fold cross validation (K-fold CV) 기법과 grid-search cross validation (Grid-search CV) 기법을 적용하였다. K-fold CV는 데이터를 k개로 분할하고 k개의 모델을 만들어 k-1개의 분할에서 훈련하고 나머지 분할에서 평가하는 방법이다.
XGB 또한 K-fold CV와 Grid-search CV를 적용하였다. XGB는 조정할 수 있는 파라미터의 종류가 매우 다양해서 조정의 범위가 넓었고 본 연구에서는 10-fold CV와 Grid-Search CV 기법을 적용하여 subsample, max_depth, colsample_bytree, learning_rate, nthread, n_estimators, min_child_weight의 하이퍼 파라미터를 선정하여 최종 모델을 구축하였다.

참고문헌 (46)

Azhari, A., N.D.A. Halim, A.A.A. Mohtar, K. Aiyub, M.T. Latif, and M. Ketzel, 2021. Evaluation and prediction of PM 10 and PM 2.5 from road source emissions in Kuala Lumpur City Centre, Sustainability, 13(10): 5402. https://doi.org/10.3390/su13105402

상세보기
Berrocal, V.J., Y. Guan, A. Muyskens, H. Wang, B.J. Reich, J.A. Mulholland, and H.H. Chang, 2020. A comparison of statistical and machine learning methods for creating national daily maps of ambient PM 2.5 concentration, Atmospheric Environment, 222: 117130. https://doi.org/10.1016/j.atmosenv.2019.117130

상세보기
Breiman, L., 2001. Random forests, Machine Learning, 45(1): 5-32. https://doi.org/10.1023/A:1010933404324
Byon, J-Y., S.-O. Hong, Y.-S. Park, and Y.-H. Kim, 2021. Evaluation of the Urban Heat Island Intensity in Seoul Predicted from KMA Local Analysis and Prediction System, Journal of the Korean Earth Science Society, 42(2): 135-148 (in Korean with English abstract). https://doi.org/10.5467/JKESS.2021.42.2.135

원문보기 상세보기
Chae, S., J. Shin, S. Kwon, S. Lee, S. Kang, and D. Lee, 2021. PM 10 and PM 2.5 real-time prediction models using an interpolated convolutional neural network, Scientific Reports, 11(1): 1-9. https://doi.org/10.1038/s41598-021-91253-9

상세보기
Chen, G., S. Li, L.D. Knibbs, N.A.S. Hamm, W. Cao, T. Li, J. Guo, H. Ren, M. J. Abramson, and Y. Guo, 2018a. A machine learning method to estimate PM 2.5 concentrations across China with remote sensing, meteorological and land use information, Science of the Total Environment, 636: 52-60. https://doi.org/10.1016/j.scitotenv.2018.04.251

상세보기
Chen, G., Y. Li, Y. Zhou, C. Shi, Y. Guo, and Y. Liu, 2021. The comparison of AOD-based and non-AOD prediction models for daily PM 2.5 estimation in Guangdong province, China with poor AOD coverage, Environmental Research, 195: 110735. https://doi.org/10.1016/j.envres.2021.110735

상세보기
Chen, G., Y. Wang, S. Li, W. Cao, H. Ren, L.D. Knibbs, M.J. Abramson, and Y. Guo, 2018b. Spatiotemporal patterns of PM 10 concentrations over China during 2005-2016: A satellite-based estimation using the random forests approach, Environmental Pollution, 242: 605-13. https://doi.org/10.1016/j.envpol.2018.07.012

상세보기
Chen, T. and C. Guestrin, 2016. Xgboost: A scalable tree boosting system, Proc. of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, Aug. 13-17, pp. 785-794. https://doi.org/10.1145/2939672.2939785
Choi, I., W. Lee, B. Eun, J. Heo, K-H. Chang, and J. Oh, 2022. A Study on Prediction of PM 2.5 Concentration Using DNN, Journal of Environmental Impact Assessment, 31(2): 83-94 (in Korean with English abstract). https://doi.org/10.14249/eia.2022.31.2.83

원문보기 상세보기
Choi, J.K., I.S. Choi, K.K. Cho, and S.H. Lee, 2020. Harmfulness of particulate matter in disease progression, Journal of Life Science, 30(2): 191-201 (in Korean with English abstract). https://doi.org/10.5352/JLS.2020.30.2.191

원문보기 상세보기
Choi, S.I, J. An, and Y.M. Jo, 2018. Review of Analysis Principle of Fine Dust, Korean Industrial Chemistry News, 21(2): 16-23 (in Korean with English abstract).
Guo, J., F. Xia, Y. Zhang, H. Liu, J. Li, M. Lou, J. He, Y. Yan, F. Wang, M. Min, and P. Zhai, 2017. Impact of diurnal variability and meteorological factors on the PM 2.5 -AOD relationship: Implications for PM 2.5 remote sensing, Environmental Pollution, 221: 94-104. https://doi.org/10.1016/j.envpol.2016.11.043

상세보기
Ha, J-E., H.-C. Shin, and Z.-K. Lee, 2017. Korean Text Classification Using Randomforest and XGBoost Focusing on Seoul Metropolitan Civil Complaint Data, The Journal of Bigdata, 2(2): 95-104 (in Korean with English abstract).
Hebert, J., 2016. Predicting rare failure events using classification trees on large scale manufacturing data with complex interactions, Proc. of 2016 IEEE International Conference on Big Data, Washington D.C., USA, Dec. 5-8, pp. 2024-2028. https://doi.org/10.1109/BigData.2016.7840825
Hwang, I.C., 2018. Particulate Matter Management Policy of Seoul: Achievements and Limitations, The Korea Association for Policy Studies, 27(2): 27-51 (in Korean with English abstract).
Hwang, S., T.H. Kim, M. Kim, and J. Choi, 2022. A Study on the Time Series Characteristics of High-concentration Fine Dust Generation by Local Indicator of Temporal Burstiness, Journal of the Korean Geographical Society, 57(1): 97-108 (in Korean with English abstract). https://doi.org/10.22776/kgs.2021.57.1.97
Kim, B-Y., Y.-K. Lim, and J-W. Cha, 2022. Short-term prediction of particulate matter (PM 10 and PM 2.5 ) in Seoul, South Korea using tree-based machine learning algorithms, Atmospheric Pollution Research, 13(10): 101547. https://doi.org/10.1016/j.apr.2022.101547

상세보기
Kim, E. and J. Moon, 2021. Analyzing the Temporal Pattern of Particulate Matter Emission Multipliers: Development of the Quarterly Input-Output Model, Journal of Environmental Policy and Administration, 29(2): 1-29 (in Korean with English abstract). http://dx.doi.org/10.15301/jepa.2021.29.2.1

상세보기
Kim, H., 2020. The Prediction of PM 2.5 in Seoul through XGBoost Ensemble, Journal of The Korean Data Analysis Society, 22(4): 1661-1671 (in Korean with English abstract). https://doi.org/10.37727/jkdas.2020.22.4.1661

상세보기
Kim, K., D. Lee, K.-Y. Lee, K.-H. Lee, and Y. Noh, 2016. Estimation of surface-level PM 2.5 concentration based on MODIS aerosol optical depth over Jeju, Korea, Korean Journal of Remote Sensing, 32(5): 413-421 (in Korean with English abstract). https://doi.org/10.7780/kjrs.2016.32.5.2

원문보기 상세보기
Kim, Y. and K. Chang, 2021. Comparison and analysis of prediction performance of fine particulate matter (PM 2.5 ) based on deep learning algorithm, Journal of Convergence for Information Technology, 11(3): 7-13 (in Korean with English abstract). https://doi.org/10.22156/CS4SMB.2021.11.03.007

원문보기 상세보기
Lee, D. and S. Lee, 2020. Hourly Prediction of Particulate Matter (PM 2.5 ) Concentration Using Time Series Data and Random Forest, Korea Information Processing Society-Transactions on Software and Data Engineering, 9(4): 129-36 (in Korean with English abstract). https://doi.org/10.3745/KTSDE.2020.9.4.129

원문보기 상세보기
Lee, S.-B., C.-H. Kang, D.-S. Jung, H.-J. Ko, H.-B. Kim, Y.-S. Oh, and H.-L. Kang, 2010. Composition and pollution characteristics of TSP, PM 2.5 atmospheric aerosols at Gosan site, Jeju Island, Analytical Science and Technology, 23(4): 372-382 (in Korean with English abstract).
Lim, J.-M., 2019. An Estimation Model of Fine Dust Concentration Using Meteorological Environment Data and Machine Learning, Journal of Information Technology Services, 18(1): 173-186 (in Korean with English abstract). https://doi.org/10.9716/KITS.2019.18.1.173

원문보기 상세보기
Lin, L., Y. Liang, L. Liu, Y. Zhang, D. Xie, F. Yin, and T. Ashraf, 2022. Estimating PM 2.5 Concentrations Using the Machine Learning RF-XGBoost Model in Guanzhong Urban Agglomeration, China, Remote Sensing, 14(20): 5239. https://doi.org/10.3390/rs14205239

상세보기
Joharestani M. Z., C. Cao, X. Ni, B. Bashir, and S. Talebiesfandarani, 2019. PM 2.5 prediction based on random forest, XGBoost, and deep learning using multisource remote sensing data, Atmosphere, 10(7): 373. https://doi.org/10.3390/atmos10070373

상세보기
Park, D.-U. and K.-C. Ha, 2008. Characteristics of PM 10 , PM 2.5 , CO2 and CO monitored in interiors and platforms of subway train in Seoul, Korea, Environment International, 34(5): 629-634. https://doi.org/10.1016/j.envint.2007.12.007

상세보기
Park, J.-Y., T.-Y. Kwon, and J.-Y. Lee, 2017. Estimation of surface visibility using MODIS AOD, Korean Journal of Remote Sensing, 33(2): 171-187 (in Korean with English abstract). https://doi.org/10.7780/kjrs.2017.33.2.6

원문보기 상세보기
Park, S., M. Kim, and J. Im, 2021. Estimation of Ground-level PM 10 and PM 2.5 Concentrations Using Boosting-based Machine Learning from Satellite and Numerical Weather Prediction Data, Korean Journal of Remote Sensing, 37(2): 321-335 (in Korean with English abstract). https://doi.org/10.7780/kjrs.2021.37.2.11

원문보기 상세보기
Park, S. and H. Shin, 2017. Analysis of the Factors Influencing PM 2.5 in Korea: Focusing on Seasonal Factors, Journal of Environmental Policy and Administration, 25(1): 227-248 (in Korean with English abstract). https://doi.org/10.15301/jepa.2017.25.1.227

상세보기
Peng, J., H. Han, Y. Yi, H. Huang, and L. Xie, 2022. Machine learning and deep learning modeling and simulation for predicting PM 2.5 concentrations, Chemosphere, 308: 136353. https://doi.org/10.1016/j.chemosphere.2022.136353

상세보기
Shin, D.-H. and Y.-M. Kim, 2015. The Utilization of Big Data's Disaster Management in Korea, The Journal of the Korea Contents Association, 15(2): 377-392 (in Korean with English abstract). https://doi.org/10.5392/JKCA.2015.15.02.377

원문보기 상세보기
Shogrkhodaei, S.Z., S.V. Razavi-Termeh, and A. Fathnia, 2021. Spatio-temporal modeling of PM 2.5 risk mapping using three machine learning algorithms, Environmental Pollution, 289: 117859. https://doi.org/10.1016/j.envpol.2021.117859

상세보기
Sihag, P., V. Kumar, F.R. Afghan, S.M. Pandhiani, and A. Keshavarzi, 2019. Predictive modeling of PM 2.5 using soft computing techniques: case study-Faridabad, Haryana, India, Air Quality, Atmosphere & Health, 12(12): 1511-1520. https://doi.org/10.1007/s11869-019-00755-z

상세보기
Son, S. and J. Kim, 2021. Vulnerability Assessment for Fine Particulate Matter (PM 2.5 ) in the Schools of the Seoul Metropolitan Area, Korea: Part I - Predicting Daily PM 2.5 Concentrations, Korean Journal of Remote Sensing, 37(6-2): 1881-1890 (in Korean with English abstract). https://doi.org/10.7780/kjrs.2021.37.6.2.10

원문보기 상세보기
Song, B.-G. and K.-H. Park, 2022. Analysis of PM 2.5 Pattern Considering Land Use Types and Meteorological Factors - Focused on Changwon National Industrial Complex -, Journal of the Korean Association of Geographic Information Studies, 25(2): 1-17 (in Korean with English abstract). https://doi.org/10.11108/kagis.2022.25.2.001

원문보기 상세보기
Stafoggia, M., T. Bellander, S. Bucci, M. Davoli, K. de Hoogh, F. de' Donato, C. Gariazzo, A. Lyapustin, P. Michelozzi, M. Renzi, M. Scortichini, A. Shtein, G. Viegi, I. Kloog, and J. Schwartz, 2019. Estimation of daily PM 10 and PM 2.5 concentrations in Italy, 2013-2015, using a spatiotemporal land-use random-forest model, Environment International, 124: 170-179. https://doi.org/10.1016/j.envint.2019.01.016

상세보기
Sung, S.H., S. Kim, and M.H. Ryu, 2020. A Comparative Study on the Performance of Machine Learning Models for the Prediction of Fine Dust: Focusing on Domestic and Overseas Factors, Innovation Studies, 15(4): 339-357 (in Korean with English abstract). https://doi.org/10.46251/INNOS.2020.11.15.4.339

상세보기
Xie, Y., Y. Wang, K. Zhang, W. Dong, B. Lv, and Y. Bai, 2015. Daily estimation of ground-level PM 2.5 concentrations over Beijing using 3 km resolution MODIS AOD, Environmental Science and Technology, 49(20): 12280-12288. https://doi.org/10.1021/acs.est.5b01413

상세보기
Xin, J., Y. Wang, Z. Li, P. Wang, W.M. Hao, B.L. Nordgren, S. Wang, G. Liu, L. Wang, and T. Wen. 2007. Aerosol optical depth (AOD) and Angstrom exponent of aerosols observed by the Chinese Sun Hazemeter Network from August 2004 to September 2005, Journal of Geophysical Research: Atmospheres, 112(D5). https://doi.org/10.1029/2006JD007075

상세보기
Yoo, H.-G., J.-W. Hong, J. Hong, S. Sung, E.J. Yoon, J.-H. Park, and J.-H. Lee, 2020. Impact of Meteorological Conditions on the PM 2.5 and PM 10 concentrations in Seoul, Journal of Climate Change Research, 11(5-2): 521-528 (in Korean with English abstract). https://doi.org/10.15531/ksccr.2020.11.5.521

상세보기
Yu, M., Y. Lee, and J. Yi, 2016. Flood inflow forecasting on HantanRiver reservoir by using forecasted rainfal, Journal of Korea Water Resources Association, 49(4): 327-333 (in Korean with English abstract). https://doi.org/10.3741/JKWRA.2016.49.4.327

원문보기 상세보기
Zhang, D., L. Qian, B. Mao, C. Huang, B. Huang, and Y. Si, 2018. A data-driven design for fault detection of wind turbines using random forests and XGboost, IEEE Access, 6: 21020-21031. https://doi.org/10.1109/ACCESS.2018.2818678

상세보기
Zhang, H., R.M. Hoff, and J.A. Engel-Cox, 2009. The relation between Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth and PM 2.5 over the United States: a geographical comparison by US Environmental Protection Agency regions, Journal of the Air & Waste Management Association, 59(11): 1358-1369. https://doi.org/10.3155/1047-3289.59.11.1358

상세보기
Zhang, H. and S. Kondragunta, 2021. Daily and hourly surface PM 2.5 estimation from satellite AOD, Earth and Space Science, 8(3): e2020EA001599. https://doi.org/10.1029/2020EA001599

상세보기

저자의 다른 논문 :

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증