Hotel in a Clickout: a Session-based RNN Prediction Approach

This thesis aims at analyzing and experimenting on a Session-based Recommender System problem, which dataset is provided by Trivago for the Recsys Challenge 2019. In order to predict the right hotel we develop a solution based on Recurrent Neural Networks, experimenting this recent technology in a field yet to be entirely explored. The diversity of the results obtained are interesting for analyzing how good this approach works in this field. Along with the solution described we analyze an ensemble of the results with a different algorithm(Matrix Factorization), which is discussed in another thesis. This comparison made it possible to make conclusions about how good different approaches are and why.