基于LightGBM算法的边坡稳定性预测研究

doi:10.16265/j.cnki.issn1003-3033.2022.07.1473

摘要/Abstract

摘要：

为减少边坡失稳造成的灾害和事故,提出基于轻量级梯度提升机(LightGBM)算法的边坡稳定性预测模型;将容重、黏聚力、内摩擦角、边坡角、边坡高度和孔隙压力比6个主要影响因素作为模型的输入,将边坡稳定性作为模型的输出;引入基于混淆矩阵的分类性能度量指标和被试工作特性曲线线下面积(AUC),评估模型的泛化性能。结果表明:提出的边坡稳定性预测模型能很好地描述影响因素与边坡稳定性之间复杂的非线性关系;与其他算法相比,LightGBM算法的F₁-Score和AUC分别为0.92和0.91,分别提高1.10%~61.40%和1.11%~28.17%;相较于0-均值归一化和反正切归一化,考虑正负相关性的最大值和最小值归一化更适合作为LightGBM模型的前处理方法;通过改变训练数据集长度,进行单因素分析,发现模型的泛化性能与训练数据集长度呈正相关关系。

关键词: 轻量级梯度提升机(LightGBM), 边坡稳定性, 机器学习算法, 混淆矩阵, 归一化

Abstract:

In order to reduce the disasters and accidents caused by the slope instability, a slope stability prediction model based on the LightGBM algorithm was proposed. Six main influencing factors, i.e. volume weight, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio, were taken as the input of the model, and the slope stability was taken as the output of the model. The classification performance metrics based on the confusion matrix and the area under curve (AUC) of the test performance curve were introduced to evaluate the generalization performance of this model. The results show that the proposed prediction model can well describe the complex nonlinear relationship between the influencing factors and the slope stability. Compared with other algorithms, the F₁-Score and AUC of the LightGBM algorithm are 0.92 and 0.91, respectively, increasing by 1.10%-61.40% and 1.11%-28.17%, respectively. Compared with zero-mean normalization and arctangent normalization, the normalization of the minimum-maximum values considering the positive and negative correlations is more applicable to the data preprocessing of LightGBM model. Furthermore, by changing the length of the training dataset and single factor analysis, it is found that the generalization performance of the model is positively correlated with the length of the training dataset.

Key words: light gradient boosting machine(LightGBM), slope stability, machine learning algorithm, confusion matrix, normalization

张凯, 张科. 基于LightGBM算法的边坡稳定性预测研究[J]. 中国安全科学学报, 2022, 32(7): 113-120.

ZHANG Kai, ZHANG Ke. Prediction study on slope stability based on LightGBM algorithm[J]. China Safety Science Journal, 2022, 32(7): 113-120.

图/表 13

图1

表1

图2

表2

表3

图3

图4

图5

表4

图6

表5

表6

表7

参考文献 22

[1]	罗亦泳, 姚宜斌, 张立亭, 等. 基于HIOA-MK-TCRVM算法的边坡稳定性估计模型[J]. 中国安全科学学报, 2017, 27(1):116-121. doi: 10.16265/j.cnki.issn1003-3033.2017.01.021
	LUO Yiyong, YAO Yibin, ZHANG Liting, et al. Slope stability estimation model based on HIOA-MK-TCRVM[J]. China Safety Science Journal, 2017, 27(1):116-121. doi: 10.16265/j.cnki.issn1003-3033.2017.01.021
[2]	吴孟龙, 叶义成, 胡南燕, 等. RAGA-PPC云模型在边坡稳定性评价中的应用[J]. 中国安全科学学报, 2019, 29(9):57-63. doi: 10.16265/j.cnki.issn1003-3033.2019.09.009
	WU Menglong, YE Yicheng, HU Nanyan, et al. Application of RAGA-PPC cloud model in slope stability evaluation[J]. China Safety Science Journal, 2019, 29(9):57-63. doi: 10.16265/j.cnki.issn1003-3033.2019.09.009
[3]	张豪, 罗亦泳. 基于人工免疫算法的边坡稳定性预测模型[J]. 煤炭学报, 2012, 37(6):911-917.
	ZHANG Hao, LUO Yiyong. Prediction model for slope stability based on artificial immune algorithm[J]. Journal of China Coal Society, 2012, 37(6):911-917.
[4]	杨杰, 马春辉, 程琳, 等. 高陡边坡变形及其对坝体安全稳定影响研究进展[J]. 岩土力学, 2019, 40(6):2341-2353, 2368.
	YANG Jie, MA Chunhui, CHENG Lin, et al. Research advances in the deformation of high-steep slopes and its influence on dam safety[J]. Rock and Soil Mechanics, 2019, 40(6):2341-2353,2368.
[5]	LIN Yun, ZHOU Keping, LI Jielin. Prediction of slope stability using four supervised learning methods[J]. IEEE Access, 2018, 6:31169-31 179.
[6]	LI Bo, LI Duanyou, ZHANG Zhijun, et al. Slope stability analysis based on quantum-behaved particle swarm optimization and least squares support vector machine[J]. Applied Mathematical Modelling, 2015, 39(17):5253-5264. doi: 10.1016/j.apm.2015.03.032
[7]	胡军, 董建华, 王凯凯, 等. 边坡稳定性的CPSO-BP模型研究[J]. 岩土力学, 2016, 37(增1):577-582,590.
	HU Jun, DONG Jianhua, WANG Kaikai, et al. Research on CPSO-BP model of slope stability[J]. Rock and Soil Mechanics, 2016, 37(S1):577-582,590.
[8]	HOANG N D, PHAM A D. Hybrid artificial intelligence approach based on metaheuristic and machine learning for slope stability assessment: a multinational data analysis[J]. Expert Systems With Applications, 2016, 46:60-68. doi: 10.1016/j.eswa.2015.10.020
[9]	QI Chongchong, TANG Xiaolin. Slope stability prediction using integrated metaheuristic and machine learning approaches: a comparative study[J]. Computers & Industrial Engineering, 2018, 118:112-122. doi: 10.1016/j.cie.2018.02.028
[10]	张凌凡, 陈忠辉, 周天白, 等. 基于梯度提升决策树的露天矿边坡多源信息融合与稳定性预测[J]. 煤炭学报, 2020, 45(增1):173-180.
	ZHANG Lingfan, CHEN Zhonghui, ZHOU Tianbai, et al. Multi-source information fusion and stability prediction of slope based on gradient boosting decision tree[J]. Journal of china coal society, 2020, 45(S1):173-180.
[11]	KARDANI N, ZHOU Annan, NAZEM M, et al. Improved prediction of slope stability using a hybrid stacking ensemble method based on finite element analysis and field data[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2021, 13(1):188-201. doi: 10.1016/j.jrmge.2020.05.011
[12]	KE Guolin, MENG Qi, FINLEY T, et al. LightGBM: a highly efficient gradient boosting decision tree[C]. Proceedings of the 31^st International Conference on Neural Information Processing Systems, 2017:3149-3157.
[13]	周挺, 杨军, 周强明, 等. 基于改进LightGBM的电力系统暂态稳定评估方法[J]. 电网技术, 2019, 43(6):1931-1940.
	ZHOU Ting, YANG Jun, ZHOU Qiangming, et al. Power system transient stability assessment method based on modified LightGBM[J]. Power System Technology, 2019, 43(6):1931-1940.
[14]	颜诗旋, 朱平, 刘钊. 基于改进 LightGBM 模型的汽车故障预测方法研究[J]. 汽车工程, 2020, 42(6):815- 819,825.
	YAN Shixuan, ZHU Ping, LIU Zhao. Research on vehicle fault prediction scheme based on improved LightGBM model[J]. Automotive Engineering, 2020, 42(6):815-819,825.
[15]	CAO Juan, ZHANG Zhao, TAO Fulu, et al. Identifying the contributions of multi-source data for winter wheat yield prediction in China[J]. Remote Sensing, 2020, 12(5):DOI: 10.3390/rs12050750. doi: 10.3390/rs12050750
[16]	SUN Yong, LIU Nan, KANG Xuming, et al. Rapid identification of geographical origin of sea cucumbers Apostichopus japonicus using FT-NIR coupled with light gradient boosting machine[J]. Food Control, 2021,124:DOI: 10.1016/j.foodcont.2021.107883. doi: 10.1016/j.foodcont.2021.107883
[17]	廖伟涵, 郭创新, 金宇, 等. 基于四阶段预处理与GBDT的油浸式变压器故障诊断方法[J]. 电网技术, 2019, 43(6):2195-2203.
	LIAO Weihan, GUO Chuangxin, JIN Yu, et al. Oil-immersed transformer fault diagnosis method based on four-stage preprocessing and GBDT[J]. Power System Technology, 2019, 43(6):2195-2203.
[18]	蔡畅, 黄亦翔, 邢宏文. 基于梯度提升树的飞机机身对接状态识别[J]. 浙江大学学报:工学版, 2019, 53(7):1274-1281.
	CAI Chang, HUANG Yixiang, XING Hongwen. State recognition for fuselage join based on gradient boosting tree[J]. Journal of Zhejiang University:Engineering Science, 2019, 53(7):1274-1281.
[19]	ZENG Ling, BIAN Hanbing, SHI Zhenning, et al. Forming condition of transient saturated zone and its distribution in residual slope under rainfall conditions[J]. Journal of Central South University, 2017, 24(8):1866-1880. doi: 10.1007/s11771-017-3594-6
[20]	汪洋, 王俊刚. 基于深度学习算法的铁路列车运行安全检测[J]. 中国安全科学学报, 2018, 28(增2):41-45.
	WANG Yang, WANG Jungang. Study on safety inspection of railway train operation based on deep learning algorithm[J]. China Safety Science Journal, 2018, 28(S2):41-45. doi: 10.16265/j.cnki.issn1003-3033.2018.S2.008
[21]	吴挡平, 张忠林, 曹婷婷. 基于Stacking策略的稳定性分类器组合模型研究[J]. 小型微型计算机系统, 2019, 40(5):1045-1049.
	WU Dangping, ZHANG Zhonglin, CAO Tingting. Research on stability classifier combination algorithm based on stacking strategy[J]. Journal of Chinese Computer Systems, 2019, 40(5):1045-1049.
[22]	ZHANG Ke, ZHANG Kai, CAI Chenxi, et al. Displacement prediction of step-like landslides based on feature optimization and VMD-Bi-LSTM: a case study of the Bazimen and Baishuihe landslides in the Three Gorges, China[J]. Bulletin of Engineering Geology and the Environment, 2021, 80(11):8481-8502. doi: 10.1007/s10064-021-02454-5

预测结果	真实值
预测结果	稳定	失稳
稳定	TP	FP
失稳	FN	TN

序号	γ/ (kN·m^-3)	c/ kPa	φ/ (°)	Φ/ (°)	H/ m	r_u	S
1	17.93	78.20	18.49	33.42	120.79	0.0	0
2	18.02	40.92	21.18	21.86	34.56	0.1	1
3	25.76	64.11	21.40	15.76	30.38	0.5	1
︙	︙	︙	︙	︙	︙	︙	︙
149	21.03	0.00	17.72	5.79	57.31	0.0	1
150	25.74	0.00	17.23	30.03	80.53	0.4	0

序号	γ/ (kN·m^-3)	c/ kPa	φ/ (°)	Φ/ (°)	H/ m	r_u	S
1	18.80	14.40	25.02	19.98	30.06	0.0	1
2	18.77	30.01	9.99	25.02	50.00	0.1	1
3	19.97	19.96	36.00	45.50	50.00	0.5	0
︙	︙	︙	︙	︙	︙	︙	︙
106	26.70	50.00	26.60	50.00	170.00	0.25	1
107	26.80	60.00	28.80	59.00	108.00	0.25	1

机器学习算法	F₁-Score	AUC	备注
支持向量机分类	0.79	0.82	文献[11]
K最近邻	0.57	0.78
决策树	0.57	0.73
极限梯度提升算法	0.76	0.77
随机森林	0.63	0.79
线性判别分析	0.76	0.80
多层感知人工神经网络	0.77	0.83
逻辑回归	0.80	0.83
极限随机树	0.71	0.74
高斯朴素贝叶斯	0.72	0.73
装袋算法	0.58	0.71
混合叠加集成算法	0.91	0.90
LightGBM算法	0.92	0.91	文中算法

方法编号	A	B	R	F₁-Score
1	0.92	0.94	0.90	0.92
2	0.83	0.87	0.80	0.83
3	0.78	0.81	0.75	0.78