基于改进随机森林的大坝监测数据质量评价算法
Data quality evaluation algorithm on dam monitoring based on improved random forest
潘宇 1李登华 2丁勇1
作者信息
- 1. 南京理工大学理学院,江苏南京 210094
- 2. 南京水利科学研究院,江苏 南京 210029;水利部水库大坝安全重点实验室,江苏南京 210029
- 折叠
摘要
针对大坝安全监测数据质量评价效率低下、智慧化不足等难题,为了满足大坝高频率自动化采集的实时数据质量评价需要,从准确性、完整性、时效性和连续性4 个方面出发提出了6 项评价因子及由相关评价规范构成的安全监测历史数据质量评价标准,通过基于AUC值改进的随机森林算法建立了大坝安全监测历史数据质量评价算法,并将该算法应用于新疆柳树沟面板堆石坝多年安全监测历史数据评价.结果表明:通过AUC值改进的随机森林算法优于原始算法,在特征属性数量取 3 时效果最好,测试集的泛化误差最小仅为0.019 5,平均准确率稳定在96.97%附近,10 折交叉验证平均准确率达到97.77%,证明了该算法的可行性.
Abstract
Aiming at the problems of low efficiency and insufficient intelligence of data quality evaluation in dam safety monito-ring,in order to meet the needs of real-time data quality evaluation of high-frequency automatic acquisition of dams,a quality evaluation criteria of safety monitoring data composed of six evaluation factors and related evaluation criteria from the four aspects of accuracy,integrity,timeliness and repair ability were proposed.And then a quality evaluation algorithm on historical data of dam safety monitoring was established by the improved random forest algorithm based on AUC value.The algorithm was applied to the evaluation of multi-year safety monitoring data of Liushugou concrete face rockfill dam in Xinjiang.The results showed that the random forest algorithm improved by AUC value was better than the original algorithm.When the feature attributes was 3,the effect was the best.The generalization error for the test set could reach 0.019 5,the average accuracy was stable at 96.97%,and the average accuracy of 10-fold cross validation reached 97.77%,which proved the feasibility of the new algorithm.
关键词
大坝安全监测/数据质量评价/随机森林算法/评价因子Key words
dam safety monitoring/data quality evaluation/random forest algorithm/evaluation factor引用本文复制引用
基金项目
国家重点研发计划项目(2022YFC3005502)
国家自然科学基金(51979174)
国家自然科学基金联合基金(U2040221)
中央级公益性科研院所基本科研业务费专项资金项目(Y321004)
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据