Reliability of information-theoretic displacement detection and risk classification for enhanced slope stability and safety at highway construction sites

In recent years, the increasing frequency of slope failure events, such as landslides, mudslides, ground fissures, and subsidence, has resulted in significant human and economic losses. Despite ongoing research, developing a comprehensive model that combines precise slope displacement detection with accurate risk classification has remained a challenge. This paper presents a novel, integrated algorithmic framework that enhances real-time detection and classification of slope displacements. Leveraging datasets from four highway construction sites, we introduce the Integrated Sequential Suspicious Detection (ISSD) algorithm, which processes vertical, horizontal, upward, and downward displacement data to identify anomalies, cluster them, and classify risks based on severity levels. The ISSD algorithm demonstrates substantial improvements over the Bregman Bubble Clustering (BBC) algorithm, with a 20 % higher precision, a 34 % increase in recall, a 28 % improvement in F-measure, and a 21 % enhancement in accuracy. The algorithm achieved training loss = 0.00, validation loss = 0.00, and accuracy = 0.98 in performance evaluation. Field validation tests further confirmed the ISSD algorithm's reliability as a decision-support tool for proactive slope failure management. This approach represents a significant advancement in safety engineering. Integrating real-time data analysis with dynamic risk classification provides a scalable and reliable solution for mitigating slope stability risks.