A Combined Model for Evaluating Strata Deformation Characteristics of Intertidal Zone Utilizing Improved Time-Series Clustering Method

The strata deformation characteristics are vital for environmental protection, geological hazard monitoring, and intertidal zone prediction. However, the study of strata deformation characteristics in intertidal zones remains a significant challenge, primarily due to the highly complex and dynamic environment, coupled with the difficulty of acquiring high-precision deformation time-series data. The main purpose of this article is to develop and verify a combined model for strata deformation considering stratified characteristics. First, a monitoring system based on a sensor array and its deployment device are developed to obtain accurate time series of strata deformation. Then, a clustering model based on agglomerative hierarchical clustering (AHC) is established, and the sensor nodes are clustered into three zones according to the similarity indicator of dynamic time warping (DTW) between their measured deformation time series to reflect the similarities and differences of deformation characteristics in different depths. The principal component analysis (PCA) is applied to reduce dimension and extract signature sequences for clusters. Based on the fivefoldcross-validation method, the prediction performances of four models are compared in the deformation time series of the clusters, including support vector regression (SVR), linear regression (LR), long short-term memory (LSTM), and informer. The prediction results show that SVR has a better prediction performance, and using feature sequences as input can effectively improve the prediction accuracy. The experiment results demonstrate that the combined model can explore the stratified deformation characteristics of each sensor node and has higher deformation prediction accuracy, making it an effective analysis method for strata deformation of intertidal zones.