基于主成分分析和长短期记忆网络的滑坡地表位移监测数据缺失插补算法

在滑坡地表位移监测过程中，由于设备工作异常或恶劣气候的干扰，原始数据会随机出现长时间序列的缺失，这类数据对滑坡的预警和预测有很大的影响。针对上述问题，提出了一种基于主成分分析(principal component analysis, PCA)和长短期记忆网络(long-short term memory, LSTM)的数据插补方法。首先利用PCA实现滑坡监测数据的降维和特征提取，消除数据间的相关性，然后建立基于LSTM的地表位移监测数据插补模型，对缺失数据进行插补。实验结果表明：该模型与BP(back propagation)神经网络等其他几种机器学习插补模型相比，平均绝对误差、均方根误差和平均绝对百分比误差分别为0.523、1.233和0.009,均优于其他几种模型；该模型能够较好地解决地表位移长时间序列数据缺失的问题。

Research on Missing Interpolation Algorithm of Landslide Surface Displacement Monitoring Data Based on Principal Component Analysis and Long Short Term Memory Network

In the process of landslide surface displacement monitoring, the original data may randomly have long time series missing due to the interference of abnormal equipment work or bad weather, and such data has a great impact on on the landslide warning and prediction. To solve these problems, this paper proposes a data interpolation method based on principal component analysis (PCA) and long short-term memory (LSTM). Firstly, PCA is used to realize the downscaling and feature extraction of landslide monitoring data to eliminate the correlation between data, and then an LSTM-based interpolation model of surface displacement monitoring data is established to interpolate the missing data; finally, the interpolation of surface displacement data is realized. The experimental results show that this model outperforms other machine learning interpolation models such as Back Propagation neural network with mean absolute error, root mean square error and mean absolute percentage error of 0.523, 1.233 and 0.009, respectively. This model can better solve the problem of missing long time series data of surface displacement.