| 摘要: |
| 堰塞坝是一种由崩塌、滑坡、泥石流等产生的斜坡体堵塞河道形成的自然灾害,严重威胁下游居民的生命财产安全。快速准确预测堰塞坝稳定性是制定灾后应急预案、开展后续重建工作的重要基础。本文通过收集国内外380座实际堰塞坝案例的数据,筛选出46组完整案例,对选取案例的数据进行相关性分析和正态转化,采用贝叶斯网络、决策树、Bagging、RBF神经网络和逻辑回归五种机器学习算法,分别建立稳定性评估模型。为对比模型拟合能力和泛化能力的优劣,比较各模型训练集和测试集的正确率及模型综合效果指标,结果显示贝叶斯网络模型正确率最高、模型效果综合指标值最大。 |
| 关键词: 堰塞坝 稳定性 机器学习 模型对比 贝叶斯网络 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金区域创新发展联合(U20A20111)；国家自然科学基金资助项目(42107189) |
|
| The stability evaluation and comparison of Landslide Dams based on Machine Learning |
|
wangdanyan1, yangxingguo2, zhoujiawen2, fengzhenyu1, liaohaimei1
|
|
1.College of Civil Engineering, Guizhou University;2.State Key Laboratory of Hydraulics and Mountain River Engineering, Sichuan University
|
| Abstract: |
| The Landslide dam is a natural disaster caused by the blockage of river channels by slope bodies caused by collapses、landslides、mudslides, etc, which seriously threatens the safety of downstream residents’ lives and property. It is important to set emergency plans and subsequent construction to establish a fast and accurate stability prediction model. This article demonstrates a database with 380 real landslide dam cases, it selects 46 cases with complete information. The statistics of the cases selected were performed through correlation analysis and normal transformation. Five stability evaluation models were created based on machine learning algorithms: Bayesian network, Decision tree, Bagging, RBF neural network, and Logistic regression. Comparing accuracy and model comprehensive indicators of training and testing sets of five models, the Bayesian network model has the biggest accuracy and comprehensive indicators. |
| Key words: landslide dam stability Machine Learning model comparison Bayesian Network |
