A Dynamic Optimization Method for High Slope Monitoring Points Based on Coupled EXDBSCAN-ATTCNN-LSTM

High slope stability monitoring is critical for infrastructure safety, yet traditional methods face challenges in adaptability and precision. This study proposes a novel dynamic optimization framework integrating Extended Density-Based Spatial Clustering (EXDBSCAN), Attention-based Convolutional Neural Networks (ATTCNN), and Long Short-Term Memory (LSTM) networks. The method enhances monitoring point layout by (1) fusing multi-source data (displacement, stress, geological parameters), (2) dynamically partitioning slope deformation zones via EXDBSCAN, and (3) optimizing monitoring points using a hybrid deep learning model. Validated through a decade-long case study of a large-scale slope along the Yellow River, the approach improved prediction accuracy from 87% to 91% after optimization, demonstrating significant advancements in spatialtemporal feature extraction, adaptive resource allocation, and early risk identification.