Recommender Systems are information filtering engines used to estimate user preferences on items they have not seen: books, movies, restaurants or other things for which individuals have dierent tastes. Collaborative and Content-based Filtering have been the two popular memory-based methods to retrieve recommendations but these suer from some limitations and might fail to provide eective recommendations. In this project we present several variations of Artificial Neural Networks, and in particular, of Autoencoders to generate model-based predictions for the users. We empirically show that a hybrid approach combining this model with other filtering engines provides a promising solution when compared to a standalone memory-based Collaborative Filtering Recommender. To wrap up the project, a chatbot connected to an e-commerce platform has been implemented so that, using Artificial Intelligence, it can retrieve recommendations to users.
Full work available here.
This project has been licensed under a Creative Commons Attribution 3.0 Unported License. Please refer to the LICENSE file on the root of this repository. Copyright (c) 2018 Laia Hedwig Esquerrà Schaefer