Ostvarenje sustava preporučivanja koristeći algoritam matrične faktorizacije

master's thesis

• Title (english): Recommendation Systems Implementation Using Matrix Factorization Algorithm
• Author: Fran Vrankić
• Mentor: Klemo Vladimir (mentor)
• Committee member: Klemo Vladimir (predsjednik povjerenstva)
• Committee member: Goran Delač (član povjerenstva)
• Committee member: Marin Šilić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2025-02-26, Croatia

Abstract

Ovaj rad proučava sustave preporučivanja, od njihovih varijanti i metoda do primjene i izazova s kojima se suočavaju. Objašnjeno je filtriranje sadržaja preko implicitnih povratnih informacija korisnika i suradničko filtriranje s pomoću eksplicitnih ocjena korisnika. Poseban je naglasak stavljen na algoritme matrične faktorizacije kao napredniju metodu koja daje kvalitetnije preporuke s većom pouzdanošću. U praktičnom dijelu rada implementiran je sustav preporučivanja koristeći osnovne metode suradničkog filtriranja i sustav koji koristi matričnu faktorizaciju. Sustavi su testirani nad skupom podataka "MovieLens", a rezultati uspoređeni s referentnom "Python" knjižnicom "Surprise". Rezultati su zadovoljavajući i pokazuju superiornost matrične faktorizacije nad običnim metodama filtracije preporuka. Rad potvrđuje važnost matrične faktorizacije u suvremenim sustavima preporučivanja te ukazuje na moguća poboljšanja (naprednije metode matrične faktorizacije).

Abstract (english)

This paper examines recommender systems, covering their variations, methods, applications, and the challenges they face. Content filtering based on users' implicit feedback and collaborative filtering using explicit user ratings are explained. Special emphasis is placed on matrix factorization algorithms as a more advanced method that provides higher-quality recommendations with greater reliability. In the practical part of the paper, a recommender system was implemented using basic collaborative filtering methods, along with a system utilizing matrix factorization. The systems were tested on the "MovieLens" dataset, and the results were compared with the reference "Python" library "Surprise". The results are satisfactory and demonstrate the superiority of matrix factorization over standard recommendation filtering methods. The paper confirms the importance of matrix factorization in modern recommender systems and highlights potential improvements (more advanced matrix factorization methods).

• Keywords: sustav preporučivanja; suradničko filtriranje; matrična faktorizacija; MovieLens; Surprise
• Keywords (english): Recommender System; Collaborative filtering; Matrix Factorization; MovieLens; Surprise
• Language: croatian
• URN:NBN: urn:nbn:hr:168:233206
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2025-03-12 11:20:09