摘要
随着近年来极端天气事件的增多,西南山区频繁遭受滑坡灾害的威胁,迫切需要建立有效的滑坡危险性评估模型以预测潜在的滑坡区域,从而降低灾害损失。然而,现有评估模型需要通过正负样本进行最佳参数求解,负样本的主观性和随机性导致模型精度受损,因此以风险区预测密度作为模型的负向平衡因子,构建一种仅利用正样本的滑坡危险性评估模型。由于模型仅包含正样本,为获得模型各输入特征最优权重,引入了人工蜂群(ABC)算法进行模型构建,并与传统的逻辑回归、支持向量机(SVM)、XGBoost和K最近邻(KNN)等算法进行对比分析。结果表明,基于正样本的评价模型在相同的风险区预测密度下预测精度更高,而相同的预测精度下风险区预测密度更低。该研究所提出的模型在西南山区滑坡灾害预警方面具有较高的精度和实际应用价值。
With the increasing frequency of extreme weather events,the southwest mountainous area faces persistent landslide threats,necessitating an effective landslide hazard assessment model to predict potential hazard areas and mitigate losses.However,existing models require both positive and negative samples for optimal parameter solving,where the subjectivity and randomness of negative samples compromise model accuracy.To address this,the authors propose a landslide hazard assessment model using only positive samples,with risk prediction density as a negative balancing factor.Since the model exclusively employs positive samples,they introduce the Artificial Bee Colony(ABC)algorithm to optimize feature weights during model construction,comparing its performance against traditional algorithms including logistic regression,SVM,XGBoost,and KNN.Results demonstrate that the positive-sample-based model achieves higher prediction accuracy at identical risk zone densities,whereas it maintains lower risk zone densities at equivalent accuracy levels.This result indicates that the model proposed in this study has high accuracy and practical application value in landslide disaster warning for southwest mountainous area.
作者
郭之政
陈军
郭海燕
吴亚平
Guo Zhizheng;Chen Jun;Guo Haiyan;Wu Yaping(Guangdong Province Surveying and Mapping Engineering Co.,Ltd.,Guangzhou 510500,China;College of Resources and Environment,Chengdu University of Information Science and Technology,Chengdu 610225,China;Sichuan Meteorological Training Center,Chengdu 610072,China;Ya'an Meteorological Bureau,Ya'an 625099,China)
出处
《资源环境与工程》
2025年第6期744-750,共7页
Resources Environment & Engineering
基金
川西南(雅安)暴雨实验室科技发展基金重大专项子项目“川西南山洪地质灾害精细化气象预警关键技术研究”(CXNBYSYSZD202402)。
关键词
滑坡
正样本
机器学习
危险性评估
模型约束
西南山区
landslide
positive sample
machine learning
hazard assessment
model constraints
south st mountainous area
