Recommender systems have become a key technology to help the users in interacting with the increasingly larger data and information available online. The rapid advancements in Deep Learning techniques have been very useful in recommendation systems as it enhances the overall performance and accuracy of the recommendation systems. This paper attempts to work on a hybrid recommendation model by considering a weighted average of top N recommendations from both content based and collaborative based filtering methods and hence eliminating their individual shortcomings. A LightFM module has been also used to evaluate the loss functions on this hybrid model and to capture the latent features about attributes of users and items. Thereafter, a class of two-layer undirected graphical models, called Restricted Boltzmann Machine (RBM) and Auto-encoder is successfully applied to the Movielens data set to provide the accurate recommendations. This study shows that the proposed approach outperform the traditional recommender systems in terms of accuracy.