作者
Wang Cen,Xuehao Liu,Ao Zhang,Jiwen Ge,Xinwen Zhao
摘要
Evaluating the carrying capacity of regional geological environments is a crucial prerequisite for achieving sustainable regional development and ecological protection. However, traditional evaluation methods often rely on manual weighting and empirical judgment, resulting in insufficient accuracy and low efficiency. This paper proposes a machine learning–based method for evaluating regional geological environmental carrying capacity. First, a multidimensional evaluation index system is constructed by integrating natural geography, socioeconomic factors, and ecological environment elements. Multisource data—including remote sensing imagery, geological surveys, and monitoring station records—are collected, cleaned, normalized, and divided into sample units. Next, machine learning algorithms such as Random Forest and XGBoost are employed, combined with feature engineering, parameter optimization, and cross-validation techniques, to build a predictive model for carrying capacity. Experimental results show that the proposed method can accurately identify high, medium, and low carrying capacity zones within the target region. The model’s predictive accuracy significantly outperforms traditional fuzzy comprehensive evaluation and weighting methods, and its computational efficiency is higher. Further feature-importance analysis reveals that indicators such as slope gradient, soil type, and human activity intensity have a notable impact on carrying capacity. Finally, the model’s applicability and limitations are discussed, and it is suggested that future work could further improve evaluation performance by integrating multisource spatiotemporal data, deep learning techniques, and quantitative uncertainty analysis. The findings provide reliable technical support for regional geological environmental protection and resource planning.