Monitoring-Based Assessment of Excavation Risk Using Data-Driven Models

Deep excavations involve complex interactions among soil properties, groundwater conditions, excavation geometry, support systems, and construction-induced responses, making reliable risk assessment essential for safe construction. With the increasing availability of monitoring data, there is a growing need for practical approaches that can systematically utilize such information for excavation risk evaluation. This study presents a monitoring-based assessment of excavation risk using data-driven modeling techniques suitable for routine geotechnical engineering practice. A comprehensive dataset comprising geotechnical, environmental, excavation-related, and structural response parameters is analyzed to classify excavation risk levels. Conventional machine learning models are employed to capture relationships among influencing factors, with emphasis placed on robustness and engineering interpretability rather than methodological complexity. Model performance is evaluated using standard classification metrics, and feature importance and sensitivity analyses are conducted to identify dominant contributors to excavation risk. Results indicate that excavation depth and deformation-related indicators govern risk classification, while environmental factors act as secondary modifiers under typical conditions. The findings demonstrate that data-driven interpretation of monitoring data can effectively complement traditional geotechnical assessment methods and support practical excavation risk evaluation.