Spatial landslide susceptibility assessment using machine learning techniques assisted by additional data created with generative adversarial networks

In recent years,landslide susceptibility mapping has substantially improved with advances in machine learning.However,there are still challenges remain in landslide mapping due to the availability of limited inventory data.In this paper,a novel method that improves the performance of machine learning techniques is presented.The proposed method creates synthetic inventory data using Generative Adversarial Networks(GANs)for improving the prediction of landslides.In this research,landslide inventory data of 156 landslide locations were identified in Cameron Highlands,Malaysia,taken from previous projects the authors worked on.Elevation,slope,aspect,plan curvature,profile curvature,total curvature,lithology,land use and land cover(LULC),distance to the road,distance to the river,stream power index(SPI),sediment transport index(STI),terrain roughness index(TRI),topographic wetness index(TWI)and vegetation density are geo-environmental factors considered in this study based on suggestions from previous works on Cameron Highlands.To show the capability of GANs in improving landslide prediction models,this study tests the proposed GAN model with benchmark models namely Artificial Neural Network(ANN),Support Vector Machine(SVM),Decision Trees(DT),Random Forest(RF)and Bagging ensemble models with ANN and SVM models.These models were validated using the area under the receiver operating characteristic curve(AUROC).The DT,RF,SVM,ANN and Bagging ensemble could achieve the AUROC values of(0.90,0.94,0.86,0.69 and 0.82)for the training;and the AUROC of(0.76,0.81,0.85,0.72 and 0.75)for the test,subsequently.When using additional samples,the same models achieved the AUROC values of(0.92,0.94,0.88,0.75 and 0.84)for the training and(0.78,0.82,0.82,0.78 and 0.80)for the test,respectively.Using the additional samples improved the test accuracy of all the models except SVM.As a result,in data-scarce environments,this research showed that utilizing GANs to generate supplementary samples is promising because it can improve the predictive capability of common landslide prediction models.     

2011~2020年勘查地球化学数据处理研究进展

我国建立了包含海量数据的高质量的勘查地球化学数据库,为矿产勘查、环境评价和地质调查等提供了重要的数据支撑。如何高效处理勘查地球化学数据,并从中发掘和识别深层次信息一直是勘查地球化学学科研究的热点和前沿领域。本文在系统调研国内外学者过去十年发表的论著基础上,对勘查地球化学数据处理方法进行分析与对比,从勘查地球化学数据库建设、地球化学异常识别及其不确定性评价等方面概述了我国近十年来在该领域取得的主要研究进展,包括:(1)分形与多重分形模型由于考虑了地球化学空间模式的复杂性和尺度不变性,在全球范围内得到极大的发展和推广,我国学者引领了基于分形与多重分形的勘查地球化学数据处理;(2)机器学习和大数据思维开始在该领域启蒙,并迅速得到关注,正在成为研究热点和前沿领域,我国学者率先开展基于机器学习算法的勘查地球化学大数据挖掘研究;(3)我国学者需要进一步加强勘查地球化学数据缺失值处理以及成分数据闭合效应研究。今后该领域应进一步加强对弱缓地球化学异常识别、异常不确定性评价以及异常识别与其形成机理相结合等方面的研究。     

基于SAPSO-ELM的瓦斯涌出量分源预测及应用

为了提高瓦斯涌出量预测精度,针对瓦斯涌出量影响因素的多重相关性、复杂性等问题,结合主成分分析法和分源预测理论,对开采层、邻近层、采空区的瓦斯涌出量数据分别进行主成分分析降维,得到预测指标。针对极限学习机(ELM)存在的输入权值矩阵与隐含层阈值随机生成的问题,利用模拟退火粒子群算法(SAPSO)对极限学习机的参数寻优,将新疆某煤矿回采工作面瓦斯涌出量及影响因素作为SAPSO-ELM模型的输入进行训练,再利用训练好的SAPSO-ELM模型对陕西某煤矿回采工作面的瓦斯涌出量进行验证预测,并对比原始ELM模型的预测结果。结果表明,SAPSO-ELM模型的平均相对误差为3.45%,ELM模型的平均相对误差为8.81%,与ELM模型相比,SAPSO-ELM模型预测精度及效率均优于原始ELM模型。分源预测理论和主成分分析法的结合有效解决了多因素间的多重相关性并降低了预测模型的复杂度,SAPSO-ELM预测模型实现了瓦斯涌出量的快速精准预测,对预防瓦斯事故发生和保障煤矿安全高效开采具有较好的指导作用。     

A novel index to evaluate discretization methods:A case study of flood susceptibility assessment based on random forest

The selection of a suitable discretization method(DM)to discretize spatially continuous variables(SCVs)is critical in ML-based natural hazard susceptibility assessment.However,few studies start to consider the influence due to the selected DMs and how to efficiently select a suitable DM for each SCV.These issues were well addressed in this study.The information loss rate(ILR),an index based on the informa-tion entropy,seems can be used to select optimal DM for each SCV.However,the ILR fails to show the actual influence of discretization because such index only considers the total amount of information of the discretized variables departing from the original SCV.Facing this issue,we propose an index,infor-mation change rate(ICR),that focuses on the changed amount of information due to the discretization based on each cell,enabling the identification of the optimal DM.We develop a case study with Random Forest(training/testing ratio of 7:3)to assess flood susceptibility in Wanan County,China.The area under the curve-based and susceptibility maps-based approaches were presented to compare the ILR and ICR.The results show the ICR-based optimal DMs are more rational than the ILR-based ones in both cases.Moreover,we observed the ILR values are unnaturally small(＜1％),whereas the ICR values are obviously more in line with general recognition(usually 10％-30％).The above results all demonstrate the superiority of the ICR.We consider this study fills up the existing research gaps,improving the ML-based natural hazard susceptibility assessments.     

矿井水防治学科基本架构及内涵

致灾机理、危险性评价、灾变预测共同构成矿井水防治理论体系基本内容,其在过去20多年里快速发展,目标是理解矿井水行为特征,预测演化趋势,服务矿区水害防治工作。机器学习是大数据时代进行数据分析和挖掘的有力工具。将机器学习应用于矿井水防治理论体系研究,已得到相对广泛的关注。针对理论体系的3项基本内容,重点讨论了机器学习在各内容建设中的具体应用,主要包括：根据不同水害类型分类简述致灾机理研究现状,指出机器学习应用暂为空白的原因为其不具备做出假设的能力。认为未来致灾机理研究方法依然以传统方法(理论分析、数值模拟、相似模拟等)为主,机器学习促进地质数据获取与处理,对机理研究作出贡献;分析方法优势,指出机器学习作用于危险性评价的主要方式为非结构化数据的处理及丰富评价方法;分析基于物理和基于数据的单一预测模式弊端,论述物理模型与数据驱动相结合的必要性,相应给出“模型−数据”双驱动预测模式的3种实现形式,并讨论了基于图像的灾变预测方法可行性。随着生产数据及地质数据的丰富,机器学习方法可推动理论体系研究快速发展,并为矿井水防治学科系统方法论研究作出贡献。

     

基于在线惯序极限学习机的瞬变电磁非线性反演

基于梯度下降法的传统人工神经网络瞬变电磁反演方法计算效率低,不能保证全局收敛。为了解决上述问题,提出一种在线惯序极限学习机(online sequential extreme learning machine, OSELM)的瞬变电磁反演方法。该方法针对瞬变电磁法所获取的高维勘探数据进行建模反演,首先,通过随机设定隐层参数(输入权值和偏差)来简化模型的学习过程;然后,将测试得到的预测样本加入训练样本中,作为下一次的更新信息,建立在线贯序极限学习机预测模型,从而最大限度提高反演精度;最后,设计了两个经典的瞬变电磁层状地电模型并进行了拟二维地电模型的反演。反演结果表明,该方法能够较好地解决瞬变电磁法高维数据非线性建模的反演问题,同时相较极限学习机(extreme learning machine, ELM),非线性反演方法具有更加准确的反演结果、更好的泛化能力以及更高的计算效率,为神经网络在地球物理反演中的应用提供了新思路。     

基于Polar1DMLP模型的CCTA冠脉管腔分割方法研究

冠状动脉数字图像造影(CCTA)是一种有效的无创评估冠脉血管狭窄等病变情况的成像技术,对CCTA的自动筛查评估依赖于冠脉管腔的高精度分割。为探索能够分割出高质量的冠脉官腔的算法,本文进行基于深度学习的端到端分割实验以及基于中心线先验信息结合CCTA灰度特征的冠脉管腔分割实验,其中基于深度学习回归方法的Polar1DMLP模型能够结合中心线先验信息得到较好的分割效果。基于公开数据集Coronary Artery Stenoses Detection and Quantification Evaluation Framework中的78组冠脉截段数据进行训练与验证,在16段数据的验证集上得到MSD(mean surface distance)为0.169 mm,DICE为0.796。结果表明本文提出的以中心线为导向信息的Polar1DMLP模型能够较好地整合血管CCTA灰度特征,回归出较为准确的冠脉血管内壁管腔轮廓半径,得到较为平滑的冠脉管腔表面模型,本方法有着较大的潜力以及拓展空间。     

Application of the numerical manifold method for stress wave propagation across rock masses

In this paper, the numerical manifold method (NMM) is extended to study wave propagation across rock masses. First, improvements to the system equations, contact treatment, and boundary conditions of the NMM are performed, where new system equations are derived based on the Newmark assumption of the space–time relationship, the edge‐to‐edge contact treatment is further developed for the NMM to handle stress wave propagation across discontinuities, and the viscous non‐reflection boundary condition is derived based on the energy minimisation principle. After the modification, numerical comparisons between the original and improved NMM are presented. The results show that the original system equations result in artificial numerical damping, which can be overcome by the Newmark system equations. Meanwhile, the original contact scheme suffers some calculation problems when modelling stress wave propagation across a discontinuity, which can be solved by the proposed edge‐to‐edge contact scheme. Subsequently, the influence of the mesh size and time step on the improved NMM for stress wave propagation is studied. Finally, 2D wave propagation is modelled, and the model's results are in good agreement with the analytical solution. Copyright © 2013 John Wiley & Sons, Ltd.     

基于集成学习的阶跃型滑坡阶跃点判别分析

针对阶跃型滑坡阶跃点识别和预测难的问题,提出了一种基于聚类分析和集成学习的阶跃型滑坡阶跃点识别和判别模型。以三峡库区八字门滑坡ZG110钻孔2010年4月至2016年12月80个滑坡位移、库水位和降雨数据为例,通过聚类分析方法识别滑坡累积位移-时间曲线中的阶跃点和平稳点,并利用K均值聚类分析检验分类结果的准确性。基于灰色关联确定了滑坡位移的最佳诱发因素,结合随机森林模型建立阶跃型滑坡阶跃点判别模型并利用八字门滑坡ZG111钻孔验证该模型的准确性。模型阶跃点和平稳点的识别准确率均达90%以上,表明该方法在阶跃型滑坡识别中具有较好的适用性,可为阶跃型滑坡的预测提供参考。     

卷积神经网络及其在矿床找矿预测中的应用——以安徽省兆吉口铅锌矿床为例

大数据人工智能地质学刚刚起步,基于大数据智能算法的地质研究是非常有意义的探索性实验。利用大数据和机器学习解决矿产预测问题,有助于人们克服不能全面考虑地质变量的困难及评估当前模型在已有数据中的可靠性。元素地表分布特征量主要受原岩成分、成矿作用影响和地表过程的影响,它们携带某些指示矿体就位的信息,即矿体在地下空间就位时在地表的响应,且未在地表过程中消失。以往的地球化学勘查工作仅仅识别异常,但未能发现矿体在地表响应的成矿特征量。本文以安徽省兆吉口铅锌矿床为例,通过机器学习,利用卷积神经网络算法,不断挖掘元素Pb分布特征与矿体地下就位空间的耦合相关性。经过1000次训练后,可以得到准确率0. 93,损失率0. 28的卷积神经网络模型。这种神经网络模型就是矿体在地下就位时元素在地表分布的响应,可以用来进行矿产资源预测。应用该模型对未知区进行预测,结果显示第53号区域具有很大概率存在尚未发现的矿体。