Landslide is one of the important geophysical hazards that can cause a severe damage in the society and economy. Anthropogenic activities, on the other hand, are accelerating the probability and extent of the landslide. As a result, a proper estimation of the landslide probability is an essential step in contemporary research. The novel ensemble approach of 'Weight of Evidence (WOE)', 'Logistic Regression (LR)', 'WOE-Classification and Regression Tree (CART)', 'WOE-Multilayer perceptron (MLP)' and 'WOE-Extreme Gradient Boosting (XGBoost)' has been considered for estimating the landslide susceptibility of Kalimpong district in India. In validation datasets, the AUC values of ensembles 'MLP-WOE, CART-WOE, LR-WOE and XGBoost-WOE' are 0.924, 0.953, 0.940 and 0.944, respectively. According to its predictive abilities, the ensemble of 'LR-WOE' is the most optimum model, followed by 'XGBoost-WOE, CART-WOE and MLP-WOE'. Aside from that, the 'WOE' model was used to assess the importance of sub-parameters individually.