An interpretable model for landslide susceptibility assessment based on Optuna hyperparameter optimization and Random Forest

Geomatics Natural Hazards and Risk, Год журнала: 2024, Номер 15(1)

Опубликована: Май 11, 2024

This study proposed an interpretable model that combines Random Forest (RF), Optuna hyperparameter optimization, and SHapley Additive exPlanations (SHAP) to achieve optimal landslide susceptibility evaluation provide explanations in the northwest region of Yunnan Province China. First, inventory 4447 landslides 23 related factors was considered for assessment. Subsequently, a hyperparameter-optimized RF developed using framework training dataset generate maps. The performance models were evaluated accuracy (ACC), precision (PPV), recall (TPR), F1-score (F1), Area Under Curve (AUC) based on Receiver Operating Characteristic. Furthermore, interpretability enhanced through implementation SHAP. demonstrated outstanding test set, achieving ACC 0.7792, PPV 0.7448, TPR 0.8769, F1 0.8055, AUC 0.8387. analysis revealed elevation, population density, distance from roads, normalized difference vegetation index primary influencing occurrences area. provides comprehensive evaluating specific regions offers invaluable insights prevention management disasters.

Язык: Английский

---

Landslide susceptibility assessment using information quantity and machine learning integrated models: a case study of Sichuan province, southwestern China

Earth Science Informatics, Год журнала: 2025, Номер 18(2)

Опубликована: Янв. 18, 2025

Язык: Английский

---

Improving landslide susceptibility prediction through ensemble recursive feature elimination and meta-learning framework

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Фев. 12, 2025

Abstract Landslides pose significant threats to ecosystems, lives, and economies, particularly in the geologically fragile Sub-Himalayan region of West Bengal, India. This study enhances landslide susceptibility prediction by developing an ensemble framework integrating Recursive Feature Elimination (RFE) with meta-learning techniques. Seven advanced machine learning models- Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Extremely Randomized Trees (ET), Gradient Boosting (GB), Extreme (XGBoost), a Meta Classifier (MC) were applied using Remote Sensing GIS tools identify key landslide-conditioning factors classify zones. Model performance was assessed through metrics such as accuracy, precision, recall, F1 score, AUC ROC curve. Among models, achieved highest accuracy (0.956) (0.987), demonstrating superior predictive ability. XGBoost, RF also performed well, accuracies 0.943 values 0.987 (GB XGBoost) 0.983 (RF). (ET) exhibited (0.946) among individual models 0.985. SVM LR, while slightly less accurate (0.941 0.860, respectively), provided valuable insights, achieving 0.972 LR 0.935. The effectively delineated into five zones (very low, moderate, high, very high), high concentrated Darjeeling Kalimpong subdivisions. These are influenced intense rainfall, unstable geological structures, anthropogenic activities like deforestation urbanization. Notably, ET, RF, GB, XGBoost demonstrated efficiency feature selection, requiring fewer input variables maintaining performance. establishes benchmark for mapping, providing scalable adaptable geospatial hazard prediction. findings hold implications land-use planning, disaster management, environmental conservation vulnerable regions worldwide.

Язык: Английский

---

Evaluating landslide susceptibility: the impact of resolution and hybrid integration approaches

Geomatics Natural Hazards and Risk, Год журнала: 2024, Номер 15(1)

Опубликована: Окт. 1, 2024

Язык: Английский

---

Comprehensive review of remote sensing integration with deep learning in landslide forecasting and future directions

Natural Hazards, Год журнала: 2025, Номер unknown

Опубликована: Март 10, 2025

Язык: Английский

---

Applications and Advancements of Spaceborne InSAR in Landslide Monitoring and Susceptibility Mapping: A Systematic Review

Remote Sensing, Год журнала: 2025, Номер 17(6), С. 999 - 999

Опубликована: Март 12, 2025

Landslides pose significant threats to human safety and socio-economic development. In recent decades, interferometric synthetic aperture radar (InSAR) technology has emerged as a powerful tool for investigating landslides. This study systematically reviews the applications of spaceborne InSAR in landslide monitoring susceptibility mapping over past decade. We highlight advancements key areas, including atmospheric delay correction, 3D monitoring, failure time prediction, enhancements spatial temporal resolution, integration with other technologies like Global Navigation Satellite System (GNSS) physical models. Additionally, we summarize various application strategies mapping, identifying gap between static nature most current studies InSAR’s dynamic potential capturing deformation velocity. Future research should integrate InSAR-derived factors variables rainfall soil moisture prediction. also emphasize that further development will require more efficient SAR data management processing strategies.

Язык: Английский

---

Insights into landslide susceptibility: a comparative evaluation of multi-criteria analysis and machine learning techniques

Geomatics Natural Hazards and Risk, Год журнала: 2025, Номер 16(1)

Опубликована: Март 5, 2025

Landslides threaten communities worldwide, resulting in financial, environmental, and human losses. Although some studies have employed machine learning (ML) algorithms multi-criteria analysis (MCA) for landslide susceptibility mapping (LSM), comparative evaluations of these methods remain scarce, particularly regarding predictor importance, performance metrics, hyperparameter optimization. This research addresses gaps by comparing logistic regression (LR), random forest (RF), support vector machines (SVM), MCA, focusing on Petrópolis, Brazil. The ML models used 29 influencing factors, encompassing geographic, geological, climatic, anthropogenic variables, where feature importance tuning were applied to identify the most significant predictors. RF achieved highest performance, with an accuracy 0.94, ROC AUC 0.98, F1 score 0.94. SVM LR also performed well, AUCs 0.96 0.95 scores 0.92 0.89, respectively. Conversely, MCA showed lower results, 0.41, 0.55. We attribute RF's robustness its adaptability diverse variable types, reduced overfitting risk, high predictive accuracy. These findings underscore strength LSM highlight ML's potential urban planning mitigate risks landslide-prone areas.

Язык: Английский

---

Decoding vegetation's role in landslide susceptibility mapping: An integrated review of techniques and future directions

Biogeotechnics, Год журнала: 2023, Номер 2(1), С. 100056 - 100056

Опубликована: Ноя. 29, 2023

Rainfall-induced landslides, exacerbated by climate change, require urgent attention to identify vulnerable regions and propose effective risk mitigation measures. Extensive research underscores the significant impact of vegetation on soil properties slope stability, emphasizing necessity incorporate effects into regional landslide susceptibility mapping. This review thoroughly examines integrating mapping, encompassing qualitative, semi-quantitative, quantitative forecasting methods. It highlights importance incorporating aspects these methods for comprehensive accurate assessment. explores diverse roles in covering both aggregated impacts individual influences, including mechanical hydrological properties, as well implications evapotranspiration rainwater interception stability. While are integrated non-deterministic input layers, considered deterministic In application methods, it is noteworthy that a considerable number studies primarily concentrate impact, particularly reinforcement provided root cohesion. The also limitations future prospects. context mapping amid changing climatic conditions, data-driven techniques encounter challenges, while present their advantages. Stressing significance impacts, paper recommends influences unsaturated water characteristic curve permeability, along with pre-wetting suction due potential interception.

Язык: Английский

---

An Attribution Deep Learning Interpretation Model for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2023, Номер 61, С. 1 - 15

Опубликована: Янв. 1, 2023

Deep learning (DL) models are increasingly used for landslide susceptibility mapping (LSM) due to their higher accuracy. However, the lack of explanations influence input contributing factors by current DL models, accurately identifying cause each remains challenging. This study proposes a novel interpretable model named Deep-Attention-LSF, which assigns significance scores at local levels attributing susceptibility. considers more predict occurrence. DeepLIFT is as an attribution branching network interpreting relationship between and event. Subsequently, classification formed combining convolutional neural long short-term memory occurrence in entire area. The performance Deep-Attention-LSF tested using inventory map Three Gorges reservoir area associated maps 18 landslide-related factors. accuracy, recall, precision, F1-score our were 0.9645, 0.9583, 0.9676, 0.9522, respectively. These suggest that outperformed compared including self-attention LSM, frequency-ratio-attention bagging random subspace naive bayes tree, gradient boosting decision forest, information value enhanced C5.0 tree model. provided reasonable attributions comparison with field investigation reports four specific cases. Combining interpretation investigations can provide comprehensive evaluating landslides, providing useful tool prevention management.

Язык: Английский

---

Exploring deep learning models for roadside landslide prediction: Insights and implications from comparative analysis

Physics and Chemistry of the Earth Parts A/B/C, Год журнала: 2024, Номер unknown, С. 103741 - 103741

Опубликована: Сен. 1, 2024

Язык: Английский

---