基于三种机器学习算法的山洪灾害风险评价

doi:10.12082/dqxxkx.2019.190185

摘要/Abstract

摘要：

依据洪灾风险概念模型,从触发因子、孕灾环境和承灾体3方面选取江西省的12个洪灾风险指标,采用k近邻、随机森林、AdaBoost 3种机器学习算法构建洪灾风险评价模型。利用精度、Kappa系数、ROC曲线（AUC值）3种定量评估指标评价洪灾风险模型,基于随机森林和Boruta特征提取算法共同分析指标重要性,最后对比3种模型绘制的江西省山洪灾害风险分区图并分析山洪灾害分布特征。结果表明：① AdaBoost模型的精度、Kappa系数和AUC值的平均值为别为0.902、0.870和0.826,精度和Kappa系数略优于随机森林,AUC值与随机森林相当,而k近邻模型的3种性能指标均低于前2种算法;② 农田生产潜力、年最大6 h暴雨均值、年最大1 h暴雨均值、归一化差值植被指数、年降雨量均值这5个指标对最终的洪灾风险形成具有非常重要作用;③ 江西省较高风险区与最高风险区的面积和约占江西省总面积的34.4%,且主要分布于高降雨量、高暴雨量、农田生产潜力大的山区。

关键词: 随机森林机器学习算法, AdaBoost机器学习算法, ROC曲线, Boruta算法, 洪灾风险评价, 江西省

Abstract:

In China, floods are considered the most frequent natural disaster that can cause serious damages to the safety of human beings and severe economic losses. We chose Jiangxi Province as the study area, which frequently suffered from mountain torrents. According to the conceptual model of flood risk, 12 flood risk assessment indexes were selected from three aspects: trigger factor, hazard inducing environment, and hazard bearing agent. Three models of flood risk assessment were constructed using different machine learning algorithms, including k-Nearest Neighbor （kNN）, Random Forest （RF）, and AdaBoost. To evaluate the models' performances, we applied three quantitative performance indexes: accuracy, Kappa coefficient, and the ROC curve （AUC value）. We analyzed the importance of indexes based on Random Forest algorithm and the feature extraction algorithm of Boruta. Then, the zoning maps of mountain flood risk drawn by the three models were used to compare and analyze the pattern of mountain flood disasters. According to the outcomes of the performance analysis, the average values of accuracy, Kappa coefficient, and AUC of the AdaBoost model were 0.902, 0.870, and 0.826, respectively. The accuracy and Kappa coefficient were slightly higher than RF, the AUC value was equivalent to RF. The three performance indexes of the kNN model were all lower than those of the other two. Our findings suggest that five indexes play very important roles in the formation of the final flood disaster risk, including potential farmland productivity, average annual maximum rainstorm within six hours, average annual maximum rainstorm within one hour, NDVI, and average annual rainfall. Our mapping results show that the areas of higher and highest risk zones account for 34.4% of Jiangxi Province. The regions with higher and highest risk are mainly distributed in the vicinity of mountains with high rainfall, heavy rainstorm, and high potential of farmland production.

Key words: Random Forest Machine Learning Algorithm, AdaBoost Machine Learning Algorithm, ROC curve, Boruta algorithm, flood risk assessment, Jiangxi Province

周超, 方秀琴, 吴小君, 王雨晨. 基于三种机器学习算法的山洪灾害风险评价[J]. 地球信息科学学报, 2019, 21(11): 1679-1688.DOI:10.12082/dqxxkx.2019.190185

ZHOU Chao, FANG Xiuqin, WU Xiaojun, WANG Yuchen. Risk Assessment of Mountain Torrents based on Three Machine Learning Algorithms[J]. Journal of Geo-information Science, 2019, 21(11): 1679-1688.DOI:10.12082/dqxxkx.2019.190185

图/表 7

图1

表1

表2

图2

图3

图4

图5

参考文献 37

[1]	徐在庸 . 山洪及其防治[M]. 北京: 水利出版社, 1981.
	[ Xu Z Y. Mountain torrents and their prevention[M]. Beijing: Water Publishing, 1981. ]
[2]	黄大鹏, 刘闯, 彭顺风 . 洪灾风险评价与区划研究进展[J]. 地理科学进展, 2007,26(4):11-22. doi: 10.11820/dlkxjz.2007.04.004
	[ Huang D P, Liu C, Peng S F . Progress on assessment and regionalization of flood risk[J]. Progress in Geography, 2007,26(4):11-22. ] doi: 10.11820/dlkxjz.2007.04.004
[3]	王秋香, 崔彩霞, 姚艳丽 . 新疆不同区域洪灾受灾面积变化趋势及多尺度分析[J]. 地理学报, 2008,63(7):769-779. doi: 10.11821/xb200807011
	[ Wang Q X, Cui C X, Yao Y L . Variation trends and multi-scale analysis of flood affected area in various regions of Xinjiang[J]. Acta Geographica Sinica, 2008,63(7):769-779.] doi: 10.11821/xb200807011
[4]	González-Arqueros M L, Mendoza M E, Bocco G , et al. Flood susceptibility in rural settlements in remote zones: The case of a mountainous basin in the Sierra-Costa region of Michoacán, Mexico[J]. Journal of Environmental Management, 2018,223:685-693. doi: 10.1016/j.jenvman.2018.06.075 pmid: 29975896
[5]	Gigovic L, Pamucar D, Bajic Z , et al. Application of GIS-interval rough AHP methodology for flood hazard mapping in urban areas[J]. Water, 2017,9(6):360-386. doi: 10.1002/1520-6661(200011/12)9:6&lt;360::AID-MFM1008&gt;3.0.CO;2-Y pmid: 11243295
[6]	Tingsanchali T, Karim F . Flood-hazard assessment and risk-based zoning of a tropical flood plain: Case study of the Yom River, Thailand[J]. Hydrological Sciences Journal, 2010,55(2):145-161. doi: 10.1080/02626660903545987
[7]	Jiang W, Deng L, Chen L , et al. Risk assessment and validation of flood disaster based on fuzzy mathematics[J]. Progress in Natural Science: Materials International, 2009,19(10):1419-1425. doi: 10.1016/j.pnsc.2008.12.010
[8]	Patro S, Chatterjee C, Mohanty S , et al. Flood inundation modeling using MIKE FLOOD and remote sensing data[J]. Journal of the Indian Society of Remote Sensing, 2009,37(1):107-118. doi: 10.1007/s12524-009-0002-1
[9]	Kalantari Z, Cavalli M, Cantone C , et al. Flood probability quantification for road infrastructure: Data-driven spatial-statistical approach and case study applications[J]. Science of the Total Environment, 2017,581:386-398. doi: 10.1016/j.scitotenv.2016.12.147 pmid: 28062101
[10]	Tehrany M S, Pradhan B, Mansor S , et al. Flood susceptibility assessment using GIS-based support vector machine model with different kernel types[J]. Catena, 2015,125:91-101. doi: 10.1016/j.catena.2014.10.017
[11]	Zhao G, Pang B, Xu Z , et al. Mapping flood susceptibility in mountainous areas on a national scale in China[J]. Science of the Total Environment, 2018,615:1133-1142. doi: 10.1016/j.scitotenv.2017.10.037 pmid: 29751419
[12]	Kia M B, Pirasteh S, Pradhan B , et al. An artificial neural network model for flood simulation using GIS: Johor River Basin, Malaysia[J]. Environmental Earth Sciences, 2012,67(1):251-264. doi: 10.1007/s12665-011-1504-z
[13]	Tehrany M S, Jones S, Shabani F . Identifying the essential flood conditioning factors for flood prone area mapping using machine learning techniques[J]. Catena, 2019,175:174-192. doi: 10.1016/j.catena.2018.12.011
[14]	赖成光, 陈晓宏, 赵仕威 , 等. 基于随机森林的洪灾风险评价模型及其应用[J]. 水利学报, 2015,46(1):58-66.
	[ Lai C G, Chen X H, Zhao S W . et al. A flood risk assessment model based on Random Forest and its application[J]. Journal of Hydraulic Engineering, 2015,46(1):58-66. ]
[15]	Lee S, Kim J C, Jung H S . Spatial prediction of flood susceptibility using random-forest and boosted-tree models in Seoul metropolitan city, Korea[J]. Geomatics Natural Hazards & Risk, 2017,8(2):1185-1203. doi: 10.1007/s10661-019-7715-6 pmid: 31485854
[16]	Breiman L . Random forests[J]. Machine Learning, 2001,45(1):5-32. doi: 10.1023/A:1010933404324
[17]	Sánchez A S, Iglesias-Rodríguez F J, Fernández P R , et al. Applying the K-nearest neighbor technique to the classification of workers according to their risk of suffering musculoskeletal disorders[J]. International Journal of Industrial Ergonomics, 2016,52:92-99. doi: 10.1016/j.ergon.2015.09.012
[18]	Deng Z Y, Zhu X S, Cheng D B , et al. Efficient kNN classification algorithm for big data[J]. Neurocomputing, 2016,195:143-148. doi: 10.1155/2017/9290230 pmid: 28316616
[19]	Jiang S Y, Pang G S, Wu M L , et al. An improved K-nearest-neighbor algorithm for text categorization[J]. Expert Systems with Applications, 2012,39(1):1503-1509. doi: 10.1016/j.eswa.2011.08.040
[20]	Tobler W . A computer movie simulating urban growth in the Detroit region[J]. Economic Geography, 1970,46(2):234-240. doi: 10.2307/143141
[21]	Freund Y, Schapire R E . A decision-theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences, 1997,55(1):119-139. doi: 10.1006/jcss.1997.1504
[22]	曹莹, 苗启广, 刘家辰 , 等. AdaBoost算法研究进展与展望[J]. 自动化学报, 2013,39(6):745-758. doi: 10.3724/SP.J.1004.2013.00745
	[ Cao Y, Miao Q G, Liu J C , et al. Advance and prospects of AdaBoost algorithm[J]. Acta Automatica Sinica, 2013,39(6):745-758. ] doi: 10.3724/SP.J.1004.2013.00745
[23]	Zhao Y, Gong L, Zhou B , et al. Detecting tomatoes in greenhouse scenes by combining AdaBoost classifier and colour analysis[J]. Biosystems Engineering, 2016,148:127-137. doi: 10.1016/j.biosystemseng.2016.05.001
[24]	Nizamani S, Memon N, Wiil U K . Detection of illegitimate emails using boosting algorithm[J]. Counter Terrorism and Open Source Intelligence, 2011,2:249-264.
[25]	Rios-Cabrera R, Tuytelaars T, Van Gool L . Efficient multi-camera vehicle detection, tracking, and identify-cation in a tunnel surveillance application[J]. Computer Vision and Image Understanding, 2012,116(6):742-753. doi: 10.1016/j.cviu.2012.02.006
[26]	管珉, 陈兴旺 . 江西省山洪灾害风险区划初步研究[J]. 暴雨灾害, 2007(4):339-343.
	[ Guan M, Chen X W . Research of regional torrent risk zonation in Jiangxi Province[J]. Torrential Rain and Disasters, 2007(4):339-343. ]
[27]	Davidson R A, Lambert K B . Comparing the hurricane disaster risk of U.S. coastal counties[J]. Natural Hazards Review, 2001,2(3):132-142. doi: 10.1061/(ASCE)1527-6988(2001)2:3(132)
[28]	方秀琴, 王凯, 任立良 , 等. 基于GIS的江西省山洪灾害风险评价与分区[J]. 灾害学, 2017,32(1):111-116.
	[ Fang X Q, Wang K, Ren L L , et al. Risk assessment and zoning of mountain torrent disaster based on GIS in Jiangxi Province[J]. Journal of Catastrophology, 2017,32(1):111-116. ]
[29]	江西省水文局. 江西省暴雨洪水查算手册(2010版)[EB/OL].
	[ Hydrographic Office of Jiangxi Province. Rainstorm and flood calculation manual in Jiangxi Province (2010) [EB/OL]. . ]
[30]	中国科学院资源环境科学数据中心[DB/OL].
	[ Resource and Environment Data Cloud Platform[DB/OL]. ]
[31]	Huang J, Ling C X . Using AUC and accuracy in evaluating learning algorithms[J]. IEEE Transactions on Knowledge & Data Engineering, 2005,17(3):299-310. doi: 10.1109/TVCG.2018.2864814 pmid: 30130212
[32]	Bradley A P . The use of the area under the ROC curve in the evaluation of machine learning algorithms[J]. Pattern Recognition, 1997,30(7):1145-1159. doi: 10.1038/s41398-019-0638-8 pmid: 31740657
[33]	Monserud R A, Leemans R . Comparing global vegetation maps with the Kappa statistic[J]. Ecological Modelling, 1992,62(4):275-293. doi: 10.1016/0304-3800(92)90003-W
[34]	汪云云, 陈松灿 . 基于AUC的分类器评价和设计综述[J]. 模式识别与人工智能, 2011,24(1):64-71.
	[ Wang Y Y, Chen S C . A survey of evaluation and design for AUC based classifier[J]. Pattern Recognition and Artificial Intelligence, 2011,24(1):64-71. ]
[35]	Swets J A . Measuring the accuracy of diagnostic systems[J]. Science, 1988,240(4857):1285-1293. doi: 10.1126/science.3287615 pmid: 3287615
[36]	Freund Y, Schapire R E . Experiments with a new boosting algorithm[J]. Machine Learning: Proceedings of the Thirteenth International Conference, 1996,13:148-156. doi: 10.3390/s19204383 pmid: 31658774
[37]	Kursa M B, Rudnicki W R . Feature selection with the Boruta package[J]. Journal of Statistical Software, 2010,36(11):1-13.

因子	指标	缩写	数据来源
触发因子	年最大10 min暴雨均值/mm	exm10	江西省暴雨洪水查算手册（2010版）^[29]
	年最大1 h暴雨均值/mm	exm60
	年最大6 h暴雨均值/mm	exh6
	年降雨量均值/mm	rain
	年暴雨天数/d	rainday
孕灾环境	高程/m	dem	中国科学院资源环境科学数据中心^[30]
	归一化差值植被指数	ndvi	中国科学院资源环境科学数据中心^[30]
	到河最近距离/m	neardist	利用ArcGIS计算
	地形湿度指数	twi	利用ArcGIS计算
承灾体	人口密度/(人/km²)	pop	中国科学院资源环境科学数据中心^[30]
	GDP/万元	gdp
	农田生产潜力/(kg/ha)	fpp