逻辑回归优化的静—动态耦合模型在滑坡位移预测中的应用

Application of static-dynamic coupling model optimized by logistic regressionin landslide displacement prediction

滑坡位移的准确预测是滑坡预警系统的重要组成部分。本研究提出一种基于逻辑回归优化的静—动态耦合滑坡位移预测模型,以应对滑坡演化的动态特性体现在传统静态预测模型中的不足。以八字门滑坡为案例进行研究,首先采用移动平均法将累积位移分解为趋势项和周期项两个部分,随后采用静态机器学习算法——支持向量回归(SVR)和动态机器学习算法——长短期记忆神经网络(LSTM)来预测滑坡位移;其次,通过引入逻辑回归分类算法(LR),在原输入因子的基础上进行筛选,对SVR模型和LSTM模型的预测结果进行分类计算;最后,通过逻辑回归模型的输出,更新静动态耦合模型的结果,得到优化的SVR-LSTM滑坡位移预测模型。结果显示,优化后的模型相较于SVR模型和LSTM模型,其RMSE和MAPE分别降低了5.93 mm、0.28%和0.71 mm、0.03%。集成模型融合了静态(SVR)和动态(LSTM)模型的优势,其预测性能优于单一的SVR模型和LSTM模型。本研究为滑坡位移预测模型提供了一种新思路,可以为三峡库区的地质灾害预测提供参考。

Accurate prediction of landslide displacement is an important component of landslide earlywarning systems.In this study,a static-dynamic coupling landslide displacement prediction model based on logistic regression optimization is proposed to deal with the shortcomings of the dynamic characteristics of landslide evolution reflected in the traditional static prediction model.In this paper,the Bazimen landslide is taken as a case study.Firstly,the moving average method is used to decompose the cumulative displacement into two parts:trend term and periodic term.Static machine learning algorithm-support vector regression(SVR)and dynamic machine learning algorithm-long short-term memory neural network(LSTM)are used to predict landslide displacement.Secondly,by introducing the logistic regression classification algorithm,the prediction results of SVR models and LSTM models are classified and calculated on the basis of the original input factors.Finally,through the output of the logistic regression model,the results of the static and dynamic coupling model are updated,and the optimized SVR-LSTM landslide displacement prediction model is obtained.The results show that compared with SVR model and LSTM model,the RMSE and MAPE of the optimized model are reduced by 5.93 mm,0.28% and 0.71 mm,0.03% respectively.The integrated model combines the advantages of static(SVR)and dynamic(LSTM)models,and its prediction performance is better than that of a single SVR model and LSTM model.This study provides a new idea for the landslide displacement prediction model,which can provide a reference for the prediction of geological disasters in the Three Gorges Reservoir area.