| 院威麟,赵 岩,张世军,岳东霞.基于机器学习的白龙江流域泥石流发生频率预测研究[J].甘肃地质,2025,34(4):56-62 |
| 基于机器学习的白龙江流域泥石流发生频率预测研究 |
| PREDICTION ON DEBRIS FLOW OCCURRENCE FREQUENCY IN THE BAILONGJIANG RIVER BASIN BASED ON MACHINE LEARNING |
| |
| DOI： |
| 中文关键词: 泥石流 发生频率 预测模型 机器学习 白龙江 |
| 英文关键词:debris flow frequency predictive model machine learning Bailongjiang River |
| 基金项目:国家自然科学基金项目（42130709，42407201）；甘肃省科技重大专项计划项目（22ZD6FA051）；中央引导地方科技发展资金项目（24ZYQA046）；甘肃省技术创新中心建设项目（18JR2JA006）；甘肃省地质灾害野外科学观测研究站（20JR10RA657）。 |
|
| 摘要点击次数: 4 |
| 全文下载次数: 5 |
| 中文摘要: |
| 本文采用多种机器学习回归模型对白龙江流域泥石流发生频率进行建模和预测研究。 从激发条件、物质条件和地貌条件三个方面选取了16个影响参数，通过特征重要性分析筛选出8个核心参数。 比较了线性模型、支持向量、K近邻、决策树、随机森林和人工神经网络6种模型。 通过参数优化和模型评估，确定支持向量模型为最优预测模型，其Pearson相关系数得分为0. 74。 特征重要性分析结果表明，激发条件对泥石流发生频率的影响最为显著，其中10分钟平均降雨量的贡献度最高，其次为物质条件，地貌条件的影响相对较小。 基于最优模型对白龙江流域所有沟谷进行泥石流发生频率预测，预测结果能够较好地反映实际情况。 本研究可为区域泥石流灾害风险评估和防治工程设计提供科学依据。 |
| 英文摘要: |
| Abstract：This paper conducts a model construction and prediction study on the frequency of debris flow occurrence in the Bailong River Basin based on the Python environment and the Anaconda data science platform, using multiple machine learning regression models. Sixteen influencing parameters were selected from three aspects: triggering conditions, material conditions, and geomorphic conditions. Through feature importance analysis, eight core parameters were screened out. The prediction performance of six models, including linear regression, support vector regression, K -nearest neighbor regression, decision tree regression, random forest regression, and artificial neural network, was compared. Through parameter optimization and model evaluation, the support vector regression model was determined as the optimal prediction model, with a Pearson correlation coefficient score of 0. 74. The results show that the triggering conditions have the most significant impact on the frequency of debris flow occurrence, among which the 10-minute average rainfall has the highest contribution. The material conditions have the second most significant impact, while the geomorphic conditions have a relatively smaller influence. Based on the optimal model, the frequency of debris flow occurrence in all gullies of the Bailong River Basin was predicted, and the results show that the model can well reflect the actual distribution pattern. This study can provide a scientific basis for regional debris flow disaster risk assessment and prevention and control engineering design. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
