Hybrid Slide Sense: A Holistic Approach to Predicting Landslide Susceptibility via Deep Learning and Machine Learning

Landslide forecasting and management plans do come as a necessity mainly because such events pose an excellent threat to infrastructure and human life all over the world. This work presents a new approach in landslide susceptibility mapping by incorporating a mix of deep learning techniques and standard machine learning techniques. The paper suggested a hybrid approach. This approach considered integrating the Random Forest classifier into consideration with multi-geographical and environmental parameters together with CNN for classifying high resolution satellite imagery data. Using a hybrid method is advantageous because it captures the advantages of comprehensive feature extraction in rich spatial features by the Random Forest classifier together with robust classification from the CNN classification model, therefore implying better performance in predicting predictions. The model was trained and validated on an interdisciplinary dataset of multiple landslide and non-landslide occurrences from a geologically complicated region. Outputs indicate the performance of this integrated method as compared to standalone models therefore underlining the significance of the method in the integration of geographical and non-spatial data used in landslide prediction. Thirdly, the research identifies critical variables that control landslide susceptibility with educative values reducing the potential aspects of catastrophes happening. Land-use planning for sensitive areas and landslide risk assessment will have great potential due to system versatility and expansiveness; further research work adds to greater sophistication by integrating machine learning techniques with real-time environmental data in increasing the prediction abilities.