Dam deformation prediction model based on Blending-Clustering ensemble learning
[Objective]Deformation is the most intuitive effect size to reflect the structural properties and morphological changes of the dam.It is an important means that constructing a scientific and reasonable deformation prediction model to ensure the safe and healthy operation of the dam.Aiming at the false alarm phenomenon caused by the low prediction accuracy and high false positive rate of traditional dam deformation prediction models,[Methods]a dam deformation prediction model based on Blending Blending-Clustering ensemble learning is constructed by selecting different prediction models and clustering algorithms.The core of the model is to improve the prediction accuracy of single prediction models by Blending.The stability of the model is improved by clustering optimization prediction values by Clustering.Taking the deformation monitoring data of a faced rockfill dam in Xin-jiang as an example,the prediction accuracy and stability of the proposed model are comprehensively evaluated by comparing the prediction performance of multiple models.[Results]The result show that root mean square error(RMSE)and normalization mean absolute percentage error(nMAPE)of the Blending-clustering model are significantly reduced by the integration of the pre-diction model and Clustering algorithm and the prediction accuracy of the model is significantly improved.The regression correla-tion coefficient(R2)is improved and the model had stronger fitting ability.The fluctuation range of the prediction and evaluation indexes on the multi-point deformation data set of a faced rockfill dam is smaller,and the generalization and stability of the model are effectively enhanced.[Conclusion]The result indicate that the Blending-Clustering prediction model can significantly improve the prediction accuracy,generalization and stability,and has certain application value for practical engineering.
国家科技期刊平台
NSTL
万方数据
