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

Area of Science:

• Geosciences and Remote Sensing
• Environmental Science
• Data Science and Machine Learning

Background:

• Landslides present substantial risks to ecosystems, human lives, and economies, especially in geologically unstable regions like the Sub-Himalayan West Bengal.
• Accurate landslide susceptibility prediction is crucial for effective disaster management and land-use planning in vulnerable areas.

Purpose of the Study:

• To enhance landslide susceptibility prediction in West Bengal by developing an ensemble machine learning framework.
• To identify key landslide-conditioning factors and classify susceptibility zones using advanced computational techniques.
• To compare the performance of seven machine learning models for landslide prediction.

Main Methods:

• An ensemble framework integrating Recursive Feature Elimination (RFE) with meta-learning techniques was developed.
• Seven machine learning models (Logistic Regression, Support Vector Machine, Random Forest, Extremely Randomized Trees, Gradient Boosting, Extreme Gradient Boosting, and a Meta Classifier) were applied.
• Remote Sensing and GIS tools were utilized for data processing and analysis, with model performance evaluated using accuracy, precision, recall, F1 score, and AUC.

Main Results:

• The Meta Classifier (MC) achieved the highest accuracy (0.956) and AUC (0.987), outperforming individual models.
• Gradient Boosting (GB), XGBoost, and Random Forest (RF) also showed strong performance with high accuracy and AUC values.
• Extremely Randomized Trees (ET) demonstrated high accuracy (0.946) and AUC (0.985), with efficient feature selection capabilities.

Conclusions:

• The developed ensemble framework, particularly the Meta Classifier, provides a robust and scalable method for landslide susceptibility mapping.
• High and very high susceptibility zones were identified in Darjeeling and Kalimpong, influenced by rainfall, geology, and human activities.
• The findings offer critical insights for land-use planning, disaster mitigation, and environmental conservation in hazard-prone regions globally.