Creative Commons

This dissertation is the final part required for the completion of the MSc in Mobile and Web Computing at the International Hellenic University. The main purpose is the development of a product recommendation system based solely on implicit feedback. The proposed algorithm incorporates technologies that include collaborative filtering with matrix factorization and association rule mining. The proposed methodology implements a hybrid recommendation algorithm in such a way that is able to provide recommendations in multiple ways as well as use them to increase its accuracy. Moreover it includes implementation of methods for addressing data sparsity, an important issue for recommendation systems. In addition, it is implemented a relatively new approach to increase the accuracy of matrix factorization algorithms via initialization of factor vectors, which as far as we know is tested for the first time an implicit model-based collaborative filtering approach. The evaluation of the methodology shows that the implemented methods are promising and their implementation in real world scenarios could offer personalization and its benefits to customers and shop owners. I would also like to thank my supervisor Dr. Christos Tjortjis for the guidance and valuable ideas that was necessary during the elaboration of this dissertation.