摘要
开展基于滑坡灾害大数据的统计和分析,建立台风暴雨诱发滑坡概率预测模型,对预防台风暴雨诱发滑坡的防灾减灾具有重要的理论和实际价值。基于贝叶斯理论和增量学习理论,推导基于连续型变量的滑坡增量型贝叶斯分类预测概率模型;利用该模型对研究区已发生的539个滑坡点作为学习样本,进行时空联合预测预报;选取研究区部分滑坡作为待预测样本,检验模型的学习性能和预测精度。研究结果表明:该模型具有自我更新能力,在时间上实现了延续性,能捕捉连续变量微小变化对参数学习和更新的影响;在当前学习样本的条件下,增量型预测模型因为有检验元组部分样本的加入,使模型结构更为完整、概括性更强;该模型整体预测精度达75%,能较好的实现灾害时空联合预报。
It is of great theoretical and practical significance to utilize a large number of landslide disaster datasets for statistical analysis and to establish a probabilistic prediction model for typhoon rainstorm-induced landslides.Derivation of an incremental Bayesian classification prediction probabilistic model for landslides based on continuous type variables based on Bayesian theory and incremental learning theory.A dataset of 539 typhoon storm-type landslides from previous years in the study area was employed as study samples for joint spatial and temporal prediction forecasting.Some landslides in the study area were selected as samples to be predicted to test the learning performance and prediction accuracy of the model.The results are as follows.The model is self-updating and achieves continuity in time,capturing the effects of small changes in continuous variables on parameter learning and updating.Under the current conditions of learning samples,the incremental prediction model has a more complete and generalized model structure due to the inclusion of samples from the part of the test tuple.The overall prediction accuracy of the model reaches 75%,and it can better realize the joint forecast of disasters in time and space.
作者
朱祖腾
蓝燕金
简文彬
林昀昭
吴宜龙
ZHU Zuteng;LAN Yanjin;JIAN Wenbin;LIN Yunzhao;WU Yilong(Department of Geotechnical and Geological Engineering,Zijin School of Geology and Mining,Fuzhou University,Fuzhou,Fujian 350116,China;Key Laboratory of Geohazard Prevention of Fujian Province,Fuzhou University,Fuzhou,Fujian 350003,China)
出处
《水利与建筑工程学报》
2024年第6期181-188,共8页
Journal of Water Resources and Architectural Engineering
基金
国家自然科学基金项目(U2005205)。
关键词
影响因子
滑坡
贝叶斯理论
概率预测模型
时空联合预报
impact factor
landslide
Bayesian formulation
probabilistic prediction models
integrating time and space forecast
作者简介
朱祖腾(2001-),男,硕士研究生,研究方向为边坡工程。E-mail:1020255278@qq.com;通讯作者:简文彬(1963-),男,教授,主要从事岩土工程与工程地质研究工作。E-mail:jwb@fzu.edu.cn。
