Assessment of Landslide Susceptibility of the Wiśnickie Foothills Mts. (The Flysch Carpathians, Poland) Using Selected Machine Learning Algorithms

Landslides are well-known phenomena that cause significant changes to the relief of an area’s terrain, often causing damage technical infrastructure and loss life. One possible means reducing negative impact landslides on people’s lives or property is recognize areas prone their occurrence. The most common approach this problem preparing landslide susceptibility maps. These can factor in actual location causal relationship between selected environmental factors. Creating a classification landslide-prone challenging task when density not high area analysis large. We prepared shallow 10 m × resolution maps Wiśnickie Foothills (Western Carpathians, Poland) using eleven different machine learning algorithms derived from Python libraries Scikit-learn Imbalanced-Learn. analyzed characterized by mean 3.4 surficial (composed soils rocks) per km 2 . also compared approaches imbalanced sets data: Logistic Regression, Naive Bayes, Random Forest, AdaBoost, Bagging, ExtraTrees (Extremely Randomized Trees), Easy Ensemble, Balanced RUSBoost hybrid model combining Under Sampler Multi-layer Perceptron algorithms. factors (slope inclination aspect, distance rivers, lithology, soil type permeability, groundwater table depth, profile plan curvature, annual rainfall) were categorized divided into training (70%) testing (30%) sets. Accuracy, recall, G-mean under receiver operating curve (AUC) used validate quality models. results confirmed based decision tree classifiers suitable for found methods generate random undersampling subsets (Easy RUSBoost) ensemble (Bagging, Extra-Trees) both yield very similar test those use full data training. Relatively high-quality be obtained integrating algorithm with algorithm.