Extensive experiments to evaluate the suggested models were conducted over a period of four weeks with a specially developed Android application which displays POI (points of interest) recommendations to users in order to evaluate the new algorithms.The participants provided positive and negative feedback about the recommendations, while we recorded a rich set of sensor data from their mobile devices. During the experiments, we also examined the effect of explicit and latent contexts on the accuracy of context-aware recommender systems. The experiments show that all of our suggested models outperform traditional CARS in terms of RMSE (root mean square error) on the test set.