Thesis Details
Recommender System for Web Articles
Recommender systems for web articles are the main interest of this thesis. It explains the most popular approaches used to build these systems, proposes a neural-network-based architecture applying the Skip-gram inspired negative sampling method to the recommendation problem, implements this architecture together with several other models, using Singular value decomposition, collaborative filtering with Alternating Least Squares (ALS) algorithm and a content-based approach using the Doc2Vec algorithm to create document vectors from the obtained articles. Finally, it implements three evaluation metrics - namely the RANK metric, Recall at k and Precision at k - and compares the models with state-of-the-art. Apart from that it also gives a brief discussion on the role and purpose of these systems together with the motivation of using them.
Recommender Systems, Machine Learning, Deep Learning, Document Embedding, Collaborative Filtering, Matrix Factorization, Content-based filtering.
Hliněná Dana, doc. RNDr., Ph.D. (DMAT FEEC BUT), člen
Jaroš Jiří, doc. Ing., Ph.D. (DCSY FIT BUT), člen
Orság Filip, Ing., Ph.D. (DITS FIT BUT), člen
Rychlý Marek, RNDr., Ph.D. (DIFS FIT BUT), člen
@bachelorsthesis{FITBT22020, author = "Jan Ko\v{c}\'{i}", type = "Bachelor's thesis", title = "Recommender System for Web Articles", school = "Brno University of Technology, Faculty of Information Technology", year = 2019, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/22020/" }