Landslides are the most common geodynamic phenomenon in Slovakia and the most affected area is the north-western part. In the paper we evaluate the susceptibility of the region to landslides using statistical predictive modelling. In evaluating, we used the method of logistic regression and the method of random forests, while various types of variables affecting the models were taken into consideration – continuous, categorical ordinal and categorical nominal variables. We selected 15 variables that were used as predictors of a dependent variable (landslide susceptibility). Categories of categorical nominal variables were reclassified from the original detailed division into more general categories based on similarities of their characteristics. To measure the association between the input variables, Spearman`s rank correlation coefficient and Cramer`s V were used. Logistic regression model can be said to be conservative in the majority of the area, as it very rarely predicts the probability of landslide lower than 0.25 or higher than 0.75. The accuracy of the classification based on this threshold value was 79 %. The random forest model had AUCval = 0.967 and the accuracy of the classification was 94 %. Random forests achieved better results on all performance metrics considered. The predicted probabilities from the logistic regression model and the random forests model were divided into five categories of landslide susceptibility. In the logistic regression model, the slopes with very low susceptibility cover 17.2 % of the area, in the random forests model it is 34.8 % of the area.