Abstract In the North of the Moroccan kingdom, 60% of the territory is significantly influenced by geological risks, particularly mass movements that can impact the environment, society, and the economy. The central part of the Moroccan Rif is notably distinguished by conditions conducive to various types of mass movements stemming from geological, tectonic, climatic, and topographic anomalies. This study aims to determine the level of influence of predisposing factors for mass movement and to assess the susceptibility of the Bab Taza region to this risk, using three advanced machine learning models: Random Forest, Support Vector Classifier, and Logistic Regression. The Landslide Susceptibility Map (LSM), the main objective of this study, was evaluated using two statistical indices (Receiver Operating Curve & Precision-Recall Curves) and validated through field excursions. The two performance indices indicate that the “Random Forest & Logistic Regression” models are more powerful and capable of processing this dataset, facilitating the development of susceptibility maps, with performance rates reaching 0.95 in ROC and an estimated accuracy of 96% (PRC = 0.96). The field verification of the susceptibility maps (LSM) carried out shows that 60% of the study area is exposed to this risk. This highlights both the reliability of these models in this type of study and serves as a warning to local authorities for the protection of social fabric and economic activities, emphasizing prevention and environmental protection. Graphical Abstract This study presented the application of three different machine learning models to assess susceptibility to mass movements in one of the regions most threatened by such deformations in Morocco. The use of these algorithms was based on the application of numerous factors predisposed to these instabilities, derived from various data sources such as geological maps, satellite images, and field surveys. The approach in this process begins with an in-depth analysis of the 13 defined parameters and historical landslide points to define the interdependence between them. Additionally, the type of classification models used, distinguished by supervised classification, allowed for the deduction of the importance and contribution coefficient to the genesis of unstable masses. These values are considered the foundation for the development of landslide susceptibility maps (LSM) for the Bab Taza region. This study not only applies machine learning models but is also dedicated to a spatial analysis of the distribution of susceptibility levels at the scale of the studied area, as well as adding a verification and validation of the results obtained through field excursions to check the levels mentioned in the developed maps. Finally, the methodology adopted included a section to compare the models used, in order to rank their performance in this type of research using ROC and PRC performance evaluation indices, with, of course, field validation.
Sbihi et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: