~滑坡是常见的地质灾害之一，已成为一个威胁人类生命财产安全的全球性问题。滑坡的诱发因素多种多样，又具有突发性和随机性的特点，往往给人们带来严重的损失。开展区域滑坡空间预测研究不仅对于降低滑坡灾害带来的损失有着重要意义，还可以为区域滑坡预防工作提供科学依据。本文以陕西省城固县滑坡为例，系统研究了滑坡空间预测模型构建与参数优化方法。研究成果如下：
（1）根据研究区的滑坡和地质环境特征，选取了高程、坡度、坡向、平面曲率、剖面曲率、泥沙输移指数（STI）、水流功率指数（SPI）、地形湿度指数（TWI）、道路缓冲区、河流缓冲区、断层缓冲区、降雨、归一化植被指数（NDVI）、土壤、岩性和土地利用等16个因子作为研究区滑坡空间预测影响因子；对16个滑坡影响因子进行了多重共线性检验，结果显示各影响因子之间相互独立；采用相关属性评价（CAE）方法评价了16个滑坡影响因子的贡献度；利用确定系数法（CF）分析了滑坡与影响因子各分级的相关性，认为16个影响因子作为评价该地区滑坡敏感性科学合理。
（2）采用网格搜索方法对功能树（FT）、交替决策树（ADT）、逻辑模型树（LMT）、误差降低剪枝树（REPT）、装袋-功能树（Bag-FT）、装袋-误差降低剪枝树（Bag-REPT）、装袋-逻辑模型树（Bag-LMT）和装袋-交替决策树（Bag-ADT）等八个模型进行参数优化，对比了优化前后模型受试者工作特征曲线下面积（AUC值），结果显示参数优化后的八个模型的AUC值都有不同程度的增大，表明预测能力都有不同程度的增强。
（3）利用统计模型（CF）、机器学习单一模型（REPT、FT、LMT、ADT）、机器学习集合模型（Bag-REPT、Bag-FT、Bag-LMT、Bag-ADT）和优化模型（REPT、FT、LMT、ADT、Bag-REPT、Bag-FT、Bag-LMT、Bag-ADT）分别计算了城固县滑坡敏感性预测值，并在GIS软件中采用自然间断法划分为极高、高、中、低、极低五个敏感性分区。经过滑坡点密度和频率比值检验发现，优化模型（Bag-REPT、Bag-FT、Bag-LMT、Bag-ADT）敏感性分区结果相较于其他模型更加合理。
（4）采用受试者工作特征（ROC）曲线及其统计参数等对比验证了模型精确度和泛化能力。结果发现，经过超参数优化后的集合模型更适用于城固县滑坡灾害空间预测研究。其滑坡敏感性分区图可为研究区土地规划利用和防灾减灾等方面提供参考。
 

Landslide is one of the common geological disasters and has become a global problem that threatens the safety of human life and property. The inducing factors of landslides are various, and have the characteristics of suddenness and randomness, which often bring serious losses to people.Therefore, it is of great significance to carry out regional landslide spatial prediction research to reduce the losses caused by landslide disasters, and can also provide a scientific basis for regional landslide prevention work. This paper systematically studies the construction of landslide spatial prediction model and parameter optimization method based on machine learning. This paper systematically studies the construction of landslide spatial prediction model and the method of parameter optimization, taking the landslide in Chenggu County, Shaanxi Province as an example.The research results are as follows:
(1) According to the characteristics of landslide and geological environment in the study area, 16 factors were selected: elevation, slope, slope aspect, plane curvature, profile curvature, sediment transport index (STI), stream power index (SPI), topographical wetness index (TWI), road buffer, river buffer, fault buffer, rainfall, normalized difference vegetation index (NDVI), soil, lithology and land use as Influencing factors of landslide spatial prediction research in the study area; The multicollinearity test was carried out on 16 landslide influencing factors, and the results showed that the influencing factors were independent of each other; the correlation attribute evaluation (CAE) method was used to evaluate the contribution of 16 landslide influence factors; the correlation between landslides and each grade of impact factors was analyzed by the coefficient of certainty factor (CF) method, that 16 factors are scientific and reasonable to evaluate the landslide susceptibility in this area.
(2) In this paper, the grid search method is used for functional tree (FT), alternating decision tree (ADT), logistic model tree (LMT), reduced-error pruning tree (REPT), Bagging-function tree (Bag-FT), Bagging- reduced-error pruning tree (Bag-REPT), Bagging-logistic model tree (Bag-LMT) and Bagging-alternating decision tree (Bag-ADT) for parameter optimization, and the area under the receiver operating characteristic curve (AUC) of the model before and after optimization was compared. The results show that the AUC values of the eight models after parameter optimization have increased to varying degrees, indicating that the predictive ability has been enhanced to varying degrees.
(3) Calculated statistical models (CF), machine learning single models (REPT, FT, LMT, ADT), machine learning ensemble models (Bag-REPT, Bag-FT, Bag-LMT, Bag-ADT) and optimization models (REPT, FT) , LMT, ADT, Bag-REPT, Bag-FT, Bag-LMT, Bag-ADT) of Chenggu County landslide susceptibility prediction values. And in the GIS software, the natural discontinuity method is used to divide it into five susceptibility zones: extremely high, high, medium, low, and extremely low. After the landslide point density and frequency ratio test, it is found that the susceptibility partition results of the optimized models (Bag-REPT, Bag-FT, Bag-LMT, Bag-ADT) are more reasonable than other models.
(4) The Receiver Operating Characteristic (ROC) curve and its statistical parameters were used to compare the accuracy and generalization ability of the model. Comparing the prediction accuracy of each model, it is found that the model after hyperparameter optimization is more suitable for the spatial prediction of landslide disasters in Chenggu County. The landslide susceptibility map can provide a reference for land planning and utilization, disaster prevention and mitigation in the study area.
 

[1] 姚雄,余坤勇,刘健,杨素萍,何平,陈樟昊.基于随机森林模型的降水诱发山体滑坡空间预测技术[J].福建农林大学学报(自然科学版),2016,45(02):219-227.

[2] 谭龙,陈冠,曾润强,熊木齐,孟兴民.人工神经网络在滑坡敏感性评价中的应用[J].兰州大学学报(自然科学版),2014,50(01):15-20.

[3] 吴益平,殷坤龙,陈丽霞.滑坡空间预测数学模型的对比及其应用[J].地质科技情报,2007(06):95-100.

[4] 李利峰,杨华,张娟,刘琪芳.基于人工神经网络的区域滑坡预测研究[J].气象与环境科学,2020,43(03):65-70.

[5] 汪华斌,吴树仁,汪微波.滑坡灾害空间智能预测展望[J].地质科技情报,2008(02):17-20.

[6] 武雪玲,任福,牛瑞卿.多源数据支持下的三峡库区滑坡灾害空间智能预测[J].武汉大学学报(信息科学版),2013,38(08):963-968.

[7] Kavzoglu T , Colkesen I , Sahin E K. Machine Learning Techniques in Landslide Susceptibility Mapping: A Survey and a Case Study[M]. 2019.

[8] 吕江峰. 基于流域单元的永吉县泥石流敏感性分析[D]. 吉林大学.

[9] 郑玲静, 李秀珍, 徐瑞池. 基于斜坡单元的区域滑坡敏感性评价——以云南省小江流域为例[J]. 科学技术与工程, 2021, 21(28): 12322-12329.

[10] Li R, Wang N. Landslide susceptibility mapping for the Muchuan county (China): A comparison between bivariate statistical models (woe, ebf, and ioe) and their ensembles with logistic regression[J]. Symmetry, 2019, 11(6): 762.

[11] 李明波, 陈平, 陈植华,等. 湖南雪峰山地区降雨型滑坡灾害敏感性区划[J]. 地质与资源, 2019, 028(001):78-84.

[12] 邱维蓉,吴帮玉,潘学树,唐亚明.几种聚类优化的机器学习方法在灵台县滑坡易发性评价中的应用[J].西北地质,2020,53(01):222-233.

[13] 车文超. 基于机器学习算法的泥石流易发性评估——以吉林省永吉县为例[D]. 吉林大学.

[14] 李阳春,刘黔云,李潇,顾天红,张楠.基于机器学习的滑坡崩塌地质灾害气象风险预警研究[J].中国地质灾害与防治学报,2021,32(03):118-123.

[15] 田乃满,兰恒星,伍宇明,李郎平.人工神经网络和决策树模型在滑坡易发性分析中的性能对比[J].地球信息科学学报,2020,22(12):2304-2316.

[16] Hong H, Liu J, Bui D T, et al. Landslide susceptibility mapping using J48 Decision Tree with AdaBoost, Bagging and Rotation Forest ensembles in the Guangchang area (China)[J]. Catena, 2018, 163: 399-413.

[17] Hu X, Huang C, Mei H, et al. Landslide susceptibility mapping using an ensemble model of Bagging scheme and random subspace–based naïve Bayes tree in Zigui County of the Three Gorges Reservoir Area, China[J]. Bulletin of Engineering Geology and the Environment, 2021, 80(7): 5315-5329.

[18] Luo X, Lin F, Chen Y, et al. Coupling logistic model tree and random subspace to predict the landslide susceptibility areas with considering the uncertainty of environmental features[J]. Scientific reports, 2019, 9(1): 1-13.

[19] 余坤勇, 姚雄, 邱祈荣, 等. 基于随机森林模型的山体滑坡空间预测研究[J]. 农业机械学报, 2016, 47(10): 338-345.

[20] 黄发明, 陈佳武, 唐志鹏,等. 不同空间分辨率和训练测试集比例下的滑坡易发性预测不确定性[J]. 岩石力学与工程学报, 2021, 40(6):15.

[21] Chen W, Zhang S. GIS-based comparative study of Bayes network, Hoeffding tree and logistic model tree for landslide susceptibility modeling[J]. Catena, 2021, 203: 105344.

[22] Nguyen V T, Tran T H, Ha N A, et al. GIS based novel hybrid computational intelligence models for mapping landslide susceptibility: a case study at da lat city, Vietnam[J]. Sustainability, 2019, 11(24): 7118.

[23] Zhao X, Chen W. GIS-based evaluation of landslide susceptibility models using certainty factors and functional trees-based ensemble techniques[J]. Applied Sciences, 2019, 10(1): 16.

[24] 李莉. 基于随机森林模型的重庆市滑坡灾害的研究[D]. 重庆师范大学, 2015.

[25] 何静. 基于机器学习的滑坡灾害空间预测及风险评估[D]. 成都: 电子科技大学, 2019.

[26] 郭有金. 基于集成学习算法的西安市滑坡灾害易发性评价[D]. 西安科技大学, 2020.

[27] Pham B T, Prakash I, Singh S K, et al. Landslide susceptibility modeling using Reduced Error Pruning Trees and different ensemble techniques: Hybrid machine learning approaches[J]. Catena, 2019, 175: 203-218.

[28] Huang L, Wu Z, Huang W, et al. Identification of fusarium head blight in winter wheat ears based on fisher’s linear discriminant analysis and a support vector machine[J]. Applied Sciences, 2019, 9(18): 3894.

[29] Wang Y, Feng L, Li S, et al. A hybrid model considering spatial heterogeneity for landslide susceptibility mapping in Zhejiang Province, China[J]. Catena, 2020, 188: 104425.

[30] Yang L, Shami A. On hyperparameter optimization of machine learning algorithms: Theory and practice[J]. Neurocomputing, 2020, 415: 295-316.

[31] Tien Bui D, Pham B T, Nguyen Q P, et al. Spatial prediction of rainfall-induced shallow landslides using hybrid integration approach of Least-Squares Support Vector Machines and differential evolution optimization: a case study in Central Vietnam[J]. International Journal of Digital Earth, 2016, 9(11): 1077-1097.

[32] 李璐. 基于机器学习的滑坡地质灾害预报模型研究[D]. 西安工程大学, 2019.

[33] Balogun A L, Rezaie F, Pham Q B, et al. Spatial prediction of landslide susceptibility in western Serbia using hybrid support vector regression (SVR) with GWO, BAT and COA algorithms[J]. Geoscience Frontiers, 2021, 12(3): 101104.

[34] Chen W, Chen Y, Tsangaratos P, et al. Combining evolutionary algorithms and machine learning models in landslide susceptibility assessments[J]. Remote Sensing, 2020, 12(23): 3854.

[35] Xie W, Nie W, Saffari P, et al. Landslide hazard assessment based on Bayesian optimization–support vector machine in Nanping City, China[J]. Natural Hazards, 2021, 109(1): 931-948.

[36] Joodaki H, Gepner B, Kerrigan J. Leveraging machine learning for predicting human body model response in restraint design simulations[J]. Computer Methods in Biomechanics and Biomedical Engineering, 2021, 24(6): 597-611.

[37] Chen W, Li Y. GIS-based evaluation of landslide susceptibility using hybrid computational intelligence models[J]. Catena, 2020, 195: 104777.

[38] 方教勇. 基于GIS的清平地区滑坡分布分形特征与危险性评价研究[D]. 成都理工大学, 2018.

[39] 蒋婷, 徐珍, 吴强建中,等. DEM分辨率对地形湿度指数的影响[J]. 农技服务, 2017,34(13):166.

[40] Beven K J, Kirkby M J. A physically based, variable contributing area model of basin hydrology[J]. International Association of Scientific Hydrology Bulletin, 1979, 24(1):43-69.

[41] 袁士聪,谷甫刚.基于归一化指数(NDVI)的植被覆盖度分级研究——以贵州省为例[J].环保科技,2018,24(03):38-42.

[42] 蒋德明, 李益敏, 鲍华姝. 泸水县滑坡孕灾环境因素敏感性研究[J]. 自然灾害学报, 2016, 25(4): 109-119.

[43] Wang Q, Wang Y, Niu R, et al. Integration of information theory, K-means cluster analysis and the logistic regression model for landslide susceptibility mapping in the Three Gorges Area, China[J]. Remote Sensing, 2017, 9(9): 938.

[44] Chen X, Chen W. GIS-based landslide susceptibility assessment using optimized hybrid machine learning methods[J]. Catena, 2021, 196: 104833.

[45] 李利峰,张晓虎,邓慧琳,韩六平.基于熵指数与逻辑回归耦合模型的滑坡灾害易发性评价——以蓝田县为例[J].科学技术与工程,2020,20(14):5536-5543.

[46] Chen W, Hong H, Li S, et al. Flood susceptibility modelling using novel hybrid approach of Reduced-error pruning trees with Bagging and Random subspace ensembles[J]. Journal of Hydrology, 2019, 575:864-873.

[47] 邱海军. 区域滑坡崩塌地质灾害特征分析及其易发性和危险性评价研究[J]. 西北大学, 2012.

[48] Zhang T, Mao Z, Wang T. GIS-based evaluation of landslide susceptibility using a novel hybrid computational intelligence model on different mapping units[J]. Journal of Mountain Science, 2020, 17(12): 2929-2941.

[49] Ngo PTT, Panahi M, Khosravi K, et al. Evaluation of deep learning algorithms for national scale landslide susceptibility mapping of Iran[J]. Geoscience Frontiers, 2021, 12(2): 505-519.

[50] Arabameri A, Saha S, Roy J, et al. Landslide susceptibility evaluation and management using different machine learning methods in the Gallicash River Watershed, Iran[J]. Remote Sensing, 2020, 12(3): 475.

[51] 杨硕,李德营,严亮轩,黄元,王明哲.基于随机森林模型的乌江高陡岸坡滑坡地质灾害易发性评价[J].安全与环境工程,2021,28(04):131-138.

[52] Shortliffe E H . Model of Inexact Reasoning in Medicine - ScienceDirect[J]. Computer-Based Medical Consultations: Mycin, 1976:159-194.

[53] Heckerman D . Probabilistic Interpretations for MYCIN's Certainty Factors[J]. Machine Intelligence & Pattern Recognition, 2013, 4:167-196.

[54] 田春山,刘希林,汪佳.基于CF和Logistic回归模型的广东省地质灾害易发性评价[J].水文地质工程地质,2016,43(06):154-161+170.

[55] 杜晓晨. 基于ArcGIS和SPSS的德昌县滑坡危险性评价研究[D]. 西南科技大学, 2020.

[56] 刘光辉, 刘汉湖, 姚昆,等. 基于证据权法与确定系数法的滑坡危险区划对比分析-以虹口乡为例[J]. 物探化探计算技术, 2016, 038(006):848-853.

[57] 钟亩锋. 基于确定性系数与 Logistic 回归模型的地质灾害易发性评价[J]. 世界有色金属, 2019.

[58] 刘明学, 陈祥, 杨珊妮. 基于逻辑回归模型和确定性系数的崩滑流危险性区划[J]. 工程地质学报, 2014, 22(6): 1250-1256.

[59] Thai Pham B, Shirzadi A, Shahabi H, et al. Landslide susceptibility assessment by novel hybrid machine learning algorithms[J]. Sustainability, 2019, 11(16): 4386.

[60] Wu Y, Ke Y, Chen Z, et al. Application of alternating decision tree with AdaBoost and bagging ensembles for landslide susceptibility mapping[J]. Catena, 2020, 187: 104396.

[61] Hong H, Pradhan B, Xu C, et al. Spatial prediction of landslide hazard at the Yihuang area (China) using two-class kernel logistic regression, alternating decision tree and support vector machines[J]. Catena, 2015, 133: 266-281.

[62] Freund Y, Mason L. The alternating decision tree learning algorithm[C]//icml. 1999, 99: 124-133.

[63] Nefeslioglu H A, Sezer E, Gokceoglu C, et al. Assessment of landslide susceptibility by decision trees in the metropolitan area of Istanbul, Turkey[J]. Mathematical Problems in Engineering, 2010, 2010.

[64] Pfahringer B, Holmes G, Kirkby R. Optimizing the induction of alternating decision trees[C]//Pacific-Asia Conference on Knowledge Discovery and Data Mining. Springer, Berlin, Heidelberg, 2001: 477-487.

[65] Breiman L. Bagging predictors[J]. Machine learning, 1996, 24(2): 123-140.

[66] Sahana M, Pham B T, Shukla M, et al. Rainfall induced landslide susceptibility mapping using novel hybrid soft computing methods based on multi-layer perceptron neural network classifier[J]. Geocarto International, 2020: 1-25.

[67] 肖琦敏, 陈锐, 陈爽. 结合前馈神经网络和装袋算法的电力日负荷预测[J]. 微型电脑应用, 2021, 37(4): 160-163.

[68] Tien Bui D, Shahabi H, Omidvar E, et al. Shallow landslide prediction using a novel hybrid functional machine learning algorithm[J]. Remote Sensing, 2019, 11(8): 931.

[69] Gama J. Functional trees[J]. Machine learning, 2004, 55(3): 219-250.

[70] Saha S, Roy J, Hembram T K, et al. Comparison between Deep Learning and Tree-Based Machine Learning Approaches for Landslide Susceptibility Mapping[J]. Water, 2021, 13(19): 2664.

[71] Pham B T, Tien Bui D, Pourghasemi H R, et al. Landslide susceptibility assesssment in the Uttarakhand area (India) using GIS: a comparison study of prediction capability of naïve bayes, multilayer perceptron neural networks, and functional trees methods[J]. Theoretical and Applied Climatology, 2017, 128(1): 255-273.

[72] Karabulut E M, Ibrikci T. Effective automated prediction of vertebral column pathologies based on logistic model tree with SMOTE preprocessing[J]. Journal of medical systems, 2014, 38(5): 1-9.

[73] Quinlan J R. Simplifying decision trees[J]. International journal of man-machine studies, 1987, 27(3): 221-234.

[74] 路逸行. 决策树误差降低剪枝算法的改进研究[D]. 山东大学. 2020.

[75] Corsini A, Mulas M. Use of ROC curves for early warning of landslide displacement rates in response to precipitation (Piagneto landslide, Northern Apennines, Italy)[J]. Landslides, 2017, 14(3): 1241-1252.

[76] 王志恒, 胡卓玮, 赵文吉, 等. 基于多层感知器模型的区域滑坡敏感性评价研究——以四川低山丘陵区为例[J]. 防灾减灾工程学报, 2015 (5): 691-698.

[77] 邓会芳,郑善喜,周玉县,徐清清,郑家根. GIS支持下基于层次分析法的某地区滑坡敏感性评价[C]//.防治地灾 除险安居——浙江省地质学会2017年学术年会论文集.,2017:186-193.

[78] Pham B T, Tien Bui D, Dholakia M B, et al. A comparative study of least square support vector machines and multiclass alternating decision trees for spatial prediction of rainfall-induced landslides in a tropical cyclones area[J]. Geotechnical and Geological Engineering, 2016, 34(6): 1807-1824.

[79] Tien Bui D, Tuan T A, Klempe H, et al. Spatial prediction models for shallow landslide hazards: a comparative assessment of the efficacy of support vector machines, artificial neural networks, kernel logistic regression, and logistic model tree[J]. Landslides, 2016, 13(2): 361-378.

[80] Wang G, Lei X, Chen W, et al. Hybrid Computational Intelligence Methods for Landslide Susceptibility Mapping[J]. Symmetry, 2020, 12(3):325.