地质灾害不仅影响社会和经济发展，还影响到民众的生活和生计。西安市临潼区地貌类型多、人类工程活动强度高、地质环境脆弱，致使地质灾害频发，进行地质灾害易发性评价对于临潼区防灾减灾具有重要指导意义。

以临潼区崩滑地质灾害为研究对象，借助SPSS、ArcGIS、Jupyter Notebook软件，在崩滑地质灾害影响因素及其相关性分析的基础上，建立易发性评价指标体系；采用证据权模型计算各评价因子对崩滑地质灾害的贡献率大小，构建基于不同因子分级法的因子权重体系；将因子权重代入三种机器学习集成模型进行临潼区崩滑地质灾害易发性评价与模型综合性能分析；利用数学统计法对评价结果进行合理性检验；采用ROC曲线及交叉验证法检验各模型精度，最终对比因子分级方法差异，筛选出最优评价模型，实现临潼区崩滑地质灾害易发性的高精度评价。取得主要成果如下：

（1）确定了影响临潼区崩滑地质灾害发育的11个评价因子，包括地貌类型、坡度、坡向、曲率、地层岩性、土地利用、河流影响距离、构造影响距离、道路影响距离、归一化植被指数和年平均降雨量。

（2）建立了等间距-极端梯度提升、等间距-随机森林、等间距-袋装、证据权-极端梯度提升、证据权-随机森林和证据权-袋装6种评价模型，各模型训练集预测准确率分别为99.4%、93.3%、92.7%、96.1%、98.1%和96.8%，测试集预测准确率分别89.10%、91.30%、84.70%、91.30%、85.30%和94.10%，模型均表现出了较高的预测准确率，均符合合理性检验标准，其中基于证据权分级法的3种模型测试集与训练集预测准确率差异较小，泛化能力普遍更高。

（3）各模型ROC曲线下面积AUC值为0.940、0.948、0.933、0.946、0.927和0.939，五折交叉验证平均得分0.854、0.869、0.827、0.876、0.880和0.845，综合比对得到，因子分级方法与集成学习模型的最优组合为基于证据权分级法的极端梯度提升模型，该评价模型组合预测准确率最高，分区最合理。

（4）极低易发区、低易发区、中易发区、高易发区及极高易发区面积分别占临潼区总面积的64.78%、9.42%、6.31%、9.94%、9.55%，对应为606.37km²、88.18km²、59.02km²、93.08km²、89.39km²，极高易发区主要分布在骊山、仁宗、秦陵和穆寨四个街道办。

Geological disasters not only have an impact on social and economic development but also affect the well-being of people and their livelihoods. Due to the fragile geological environment, the high intensity of engineering activities and the multitude of landforms in Lintong District of Xi'an City, geological disasters have been occurring frequently. Conducting an assessment of the vulnerability of geological disasters in Lintong District would provide important guidance for disaster prevention and reduction.

Taking the collapse and landslide hazard in Lintong District as the research object, with the help of SPSS, ArcGIS, Jupyter Notebook software, based on the analysis of the influencing factors of landslide hazard and their correlation, the susceptibility evaluation index system was established. The weight of evidence model is used to calculate the contribution rate of each evaluation factor to collapse and landslide hazard, and the factor weight system based on different factor classification methods is constructed. The factor weights are substituted into three machine learning integration models to evaluate the susceptibility of collapse and landslide hazard in Lintong District and analyze the comprehensive performance of the model. Using mathematical statistics to test the rationality of the evaluation results ; the ROC curve and cross-validation method were used to test the accuracy of each model. Finally, the differences of factor classification methods were compared, and the optimal evaluation model was selected to achieve high-precision evaluation of geological disaster susceptibility in Lintong District. The main results are as follows :

(1) The predominant collapse and landslide hazard in Lintong District are small shallow tipping and sliding loess landslides, and they pose a substantial threat to the local residents. Landslide disasters are mainly small-scale shallow push type loess landslides, and most landslides currently have good stability. These geological disasters are mainly distributed in the middle and low mountainous areas of Lishan Mountain and the southern loess hilly area. During the period of strong modern human activity, they are relatively concentrated from June to September each year, and are significantly affected by rainfall and human engineering activities.

(2) Eleven evaluation factors affecting the development of geological disasters in Lintong District have been identified, including landform type, slope, aspect, curvature, stratigraphic lithology, land use, river impact distance, structural impact distance, road impact distance, Normalized Difference Vegetation Index (NDVI), and annual average rainfall.

(3) Six evaluation models, namely, E-XGBoost, E-RF, E-Bagging, WOE-XGBoost, WOE-RF and WOE-Bagging, were established to evaluate the vulnerability of geological disasters in Lintong District based on grid units. The prediction accuracy of each model training set was 99.4%, 93.3%, 92.7%, 96.1% 98.1% and 96.8%, with prediction accuracy rates of 89.10%, 91.30%, 84.70%, 91.30%, 85.30%, and 94.10% for the test set, respectively. These models all showed high prediction accuracy and met the rationality test criteria, indicating that the selected evaluation models all performed well and the partition results were reasonable. Among them, the three models based on the evidence weight grading method had smaller differences in prediction accuracy between the test set and the training set, and their generalization ability was generally higher.

(4) The area of extremely low susceptibility area, low susceptibility area, medium susceptibility area, high susceptibility area and extremely high susceptibility area accounted for 64.78 %, 9.42 %, 6.31 %, 9.94 % and 9.55 % of the total area of Lintong District, respectively, corresponding to 606.37 km²,88.18 km²,59.02 km²,93.08 km² and 89.39 km². The extremely high susceptibility area is mainly distributed in Lishan, Renzong, Qinling and Muzhai.

[1] 《工程地质手册》编委会. 工程地质手册（第五版）[M]. 北京: 中国工程建筑出版社, 2018.

[2] 张晓东. 基于遥感和GIS的宁夏盐池县地质灾害风险评价研究[D]. 中国地质大学(北京), 2018

[3] 何淑军. 陕西宝鸡市渭滨区地质灾害风险评估研究[D]. 中国地质科学院, 2009

[4] 李萍,叶辉,谈树成. 基于层次分析法的永德县地质灾害易发性评价[J]. 水土保持研究, 2021, 28(5): 394-399, 406.

[5] 赵帅,赵洲. 基于信息量模型的地质灾害易发性评价[J]. 水力发电, 2019, 45(3): 27-32.

[6] 孙炜锋,谭成轩,王继明,等. 陕西宝鸡地区千阳县地质灾害易发性评价[J]. 地质通报, 2008, (11): 1846-1853.

[7] 王雪冬,张超彪,王翠,等. 基于Logistic回归与随机森林的和龙市地质灾害易发性评价[J]. 吉林大学学报(地球科学版), 2022, 52(6): 1957-1970.

[8] 郭子正,殷坤龙,付圣,等. 基于GIS与WOE-BP模型的滑坡易发性评价[J]. 地球科学, 2019, 44(12): 4299-4312.

[9] 张俊. 三峡库区万州区滑坡灾害风险评估研究[D]. 中国地质大学, 2016

[10] 许冲,戴福初,姚鑫,等. 基于GIS与确定性系数分析方法的汶川地震滑坡易发性评价[J]. 工程地质学报, 2010, 18(1): 15-26.

[11] 吴常润,角媛梅,王金亮,等. 基于频率比-逻辑回归耦合模型的双柏县滑坡易发性评价[J]. 自然灾害学报, 2021, 30(4): 213-224.

[12] Du G, Zhang Y, Iqbal J, et al. Landslide susceptibility mapping using an integrated model of information value method and logistic regression in the Bailongjiang watershed, Gansu Province, China[J]. Journal of Mountain Science, 2017, 14: 249-268.

[13] Oh H J, Lee S, Hong S M. Landslide susceptibility assessment using frequency ratio technique with iterative random sampling[J]. Journal of Sensors, 2017, 2017.

[14] 余田野,贺姗,张科杰,等. 基于改进层次分析法的湖北省雷电灾害风险区划[J]. 暴雨灾害, 2023, 42(1): 88-96.

[15] 刘吉平,刘汉青,曾忠平,等. 基于粗糙集理论滑坡影响因子评价研究——以三峡库区青干河流域为例[J]. 水文地质工程地质, 2010, 37(5): 118-122, 132.

[16] 许冲. 2013年芦山M_s7.0级地震滑坡易发性快速评价方法[J]. 科技导报, 2013, 31(Z2): 15-23.

[17] 王佳佳,殷坤龙,肖莉丽. 基于GIS和信息量的滑坡灾害易发性评价——以三峡库区万州区为例[J]. 岩石力学与工程学报, 2014, 33(4): 797-808.

[18] 刘增源. 安康市汉滨区滑坡敏感性评价研究[D]. 西安科技大学, 2021

[19] 李利峰,张晓虎,邓慧琳,等. 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.

[20] 郭长宝,唐杰,吴瑞安,等. 基于证据权模型的川藏铁路加查——朗县段滑坡易发性评价[J]. 山地学报, 2019, 37(2): 240-251.

[21] 刘璐瑶,高惠瑛. 基于证据权与Logistic回归模型耦合的滑坡易发性评价[J]. 工程地质学报, 2021: 1-11.

[22] Beaty C B. Landslides and slope exposure[J]. The Journal of Geology, 1956, 64(1): 70-74.

[23] 王哲,易发成. 基于层次分析法的绵阳市地质灾害易发性评价[J]. 自然灾害学报, 2009, 18(1): 14-23.

[24] 李萍,叶辉,谈树成. 基于层次分析法的永德县地质灾害易发性评价[J]. 水土保持研究, 2021, 28(5): 394-399, 406.

[25] 张业成,张春山,张梁. 中国地质灾害系统层次分析与综合灾度计算[J]. 中国地质科学院院报, 1993, (Z1): 139-154.

[26] 李文彦,王喜乐. 频率比与信息量模型在黄土沟壑区滑坡易发性评价中的应用与比较[J]. 自然灾害学报, 2020, 29(4): 213-220.

[27] 范强,巨能攀,向喜琼,等. 证据权法在区域滑坡危险性评价中的应用——以贵州省为例[J]. 工程地质学报, 2014, 22(3): 474-481.

[28] Bai S B, Wang J, Lü G N, et al. GIS-based logistic regression for landslide susceptibility mapping of the Zhongxian segment in the Three Gorges area, China[J]. Geomorphology, 2010, 115(1-2): 23-31.

[29] Trigila A, Iadanza C, Esposito C, et al. Comparison of Logistic Regression and Random Forests techniques for shallow landslide susceptibility assessment in Giampilieri (NE Sicily, Italy)[J]. Geomorphology, 2015, 249: 119-136.

[30] 朱崇浩,张建经,马东华,等. 基于DInSAR-BP神经网络的震后区域滑坡危险性综合评价研究[J]. 工程地质学报, 2020, 28(3): 530-540.

[31] 苏高利,邓芳萍. 论基于MATLAB语言的BP神经网络的改进算法[J]. 科技通报, 2003, (2): 130-135.

[32] 黄国如,李碧琦. 基于模糊综合评价的深圳市暴雨洪涝风险评估[J]. 水资源与水工程学报, 2021, 32(1): 1-6.

[33] 陈晓利,祁生文,叶洪. 基于GIS的地震滑坡危险性的模糊综合评价研究[J]. 北京大学学报(自然科学版), 2008, (3): 434-438.

[34] 孙冉,王成都,夏哲兵,等. 基于AHP-信息量法的费县地质灾害风险评价[J]. 环境科学与技术, 2015, 38(S1): 430-435.

[35] 谭玉敏,郭栋,白冰心,等. 基于信息量模型的涪陵区地质灾害易发性评价[J]. 地球信息科学学报, 2015, 17(12): 1554-1562.

[36] 许露. 基于SVM-RFE和粒子群算法的特征选择算法研究[D]. 湖南师范大学, 2014

[37] 王念秦,郭有金,刘铁铭,等. 基于SVM-LR模型的滑坡易发性评价——以临潼区为例[J]. 科学技术与工程, 2019, 19(30): 62-69.

[38] 朱文慧,邹浩,何明明,等. 基于BP神经网络的地质灾害易发性分区方法研究——以蕲春县为例[J]. 资源环境与工程, 2021, 35(6): 840-844.

[39] 田乃满,兰恒星,伍宇明,等. 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.

[40] 林齐根,刘燕仪,刘连友,等. 支持向量机与Newmark模型结合的地震滑坡易发性评估研究[J]. 地球信息科学学报, 2017, 19(12): 1623-1633.

[41] 杨灿,刘磊磊,张遗立,等. 基于贝叶斯优化机器学习超参数的滑坡易发性评价[J]. 地质科技通报, 2022, 41(2): 228-238.

[42] 闫举生,谭建民. 基于不同因子分级法的滑坡易发性评价——以湖北远安县为例[J]. 中国地质灾害与防治学报, 2019, 30(1): 52-60.

[43] 廖芸婧. 安徽省歙县地质灾害易发性区划与气象预警研究[D]. 成都理工大学, 2015

[44] 周静静,张晓敏,赵法锁,等. 陕南秦巴山区地质灾害危险性评价研究[J]. 地质力学学报, 2019, 25(4): 544-553.

[45] Tehrany M S, Pradhan B, Jebur M N. Flood susceptibility mapping using a novel ensemble weights-of-evidence and support vector machine models in GIS[J]. Journal of hydrology, 2014, 512: 332-343.

[46] Khosravi K, Nohani E, Maroufinia E, et al. A GIS-based flood susceptibility assessment and its mapping in Iran: a comparison between frequency ratio and weights-of-evidence bivariate statistical models with multi-criteria decision-making technique[J]. Natural hazards, 2016, 83: 947-987.

[47] 武燕,吴映梅,李琛,等. 基于MAUP下川滇生态屏障区生态系统服务价值多尺度空间分异及地理探测响应[J]. 水土保持研究, 2023, 30(2): 333-342.

[48] 陈水满,赵辉龙,许震,等. 基于人工神经网络模型的福建南平市滑坡危险性评价[J]. 中国地质灾害与防治学报, 2022, 33(2): 133-140.

[49] 汤国安,刘学军,闾国年. 数字高程模型及地学分析的原理与方法[M]: 数字高程模型及地学分析的原理与方法, 2005.

[50] Dahal R K, Hasegawa S, Nonomura A, et al. Predictive modelling of rainfall-induced landslide hazard in the Lesser Himalaya of Nepal based on weights-of-evidence[J]. Geomorphology, 2008, 102(3-4): 496-510.

[51] Ghosh A, Dey P, Ghosh T. Integration of RS-GIS with Frequency Ratio, Fuzzy Logic, Logistic Regression and Decision Tree Models for Flood Susceptibility Prediction in Lower Gangetic Plain: A Study on Malda District of West Bengal, India[J]. Journal of the Indian Society of Remote Sensing, 2022, 50(9): 1725-1745.

[52] 魏建飞,马雷,赵卫东,等. 基于GIS与证据权模型的明光市地下水潜力分析[J]. 合肥工业大学学报(自然科学版), 2022, 45(3): 389-395, 405.

[53] 郭子正,殷坤龙,黄发明,等. 基于滑坡分类和加权频率比模型的滑坡易发性评价[J]. 岩石力学与工程学报, 2019, 38(2): 287-300.

[54] Regmi N R, Giardino J R, Vitek J D. Modeling susceptibility to landslides using the weight of evidence approach: Western Colorado, USA[J]. Geomorphology, 2010, 115(1-2): 172-187.

[55] Agterberg F P, Bonham-Carter G F, Cheng Q, et al. Weights of evidence modeling and weighted logistic regression for mineral potential mapping[J]. Computers in geology, 1993, 25: 13-32.

[56] Bonham‐Carter G F, Agterberg F P, Wright D F. Integration of geological datasets for gold exploration in Nova Scotia[J]. Digital geologic and geographic information systems, 1989, 10: 15-23.

[57] 张英. 基于ArcGIS的旬阳县地质灾害易发性分区评价[D]. 长安大学, 2014

[58] 李嘉昊. 基于PCA与改进AHP-CRITIC法的地质灾害风险评价研究[D]. 西北大学, 2022

[59] Chen W, Chen X, Peng J, et al. Landslide susceptibility modeling based on ANFIS with teaching-learning-based optimization and Satin bowerbird optimizer[J]. Geoscience Frontiers, 2021, 12(1): 93-107.

[60] Chen T, Guestrin C. Xgboost: A scalable tree boosting system[C]//Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 2016: 785-794.

[61] Breiman L. Random forests[J]. Machine learning, 2001, 45: 5-32.

[62] Breiman L. Bagging predictors[J]. Machine learning, 1996, 24: 123-140.

[63] Xu K, Ma X, Liu G, et al. Geological hazards susceptibility evaluation based on GA‐BPNN: A case study of Xingye County[J]. Earth and Space Science, 2022: e2019EA000929.

[64] Wang C, Wang X, Zhang H, et al. Assessment of environmental geological disaster susceptibility under a multimodel comparison to aid in the sustainable development of the regional economy[J]. Environmental Science and Pollution Research, 2023, 30(3): 6573-6591.