Matrična faktorizacija i plitki softmax model za implementaciju sustava preporučivanja

master's thesis

• Title (english): Matrix Factorization and Shallow Softmax Model for Recommendation Systems Implementation
• Author: Bartol Vinković
• Mentor: Klemo Vladimir (mentor)
• Committee member: Klemo Vladimir (predsjednik povjerenstva)
• Committee member: Marin Šilić (član povjerenstva)
• Committee member: Goran Delač (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-04, Croatia

Abstract

Sustavi preporučivanja služe kao filter korpusa artikala koji predlažu korisnicima relevantne preporuke radi poboljšanja korisničkog iskustva ili iz poslovnih razloga. Matrična faktorizacija je regresijski algoritam gdje se predviđaju ocjene i predstavlja korisnike i artikle u latentnom prostoru. Plitki softmax model je klasifikacijski algoritam na implicitnim podatcima koji uzima u obzir metapodatke korisnika prilikom filtriranja sadržaja. Matrična faktorizacija je precizna i implementacijski jednostavna ako se trenira gradijentnim spustom, ali manjka u skalabilnosti. Plitki softmax model ne funkcionira dobro na malim skupovima podataka.

Abstract (english)

Recommender systems filter a corpus of items to improve user experience by providing relevant recommendations or to improve the quality of a product. Matrix factorization is a regression algorithm that predicts ratings while representing users and items in a latent feature space. The shallow softmax model is a classification algorithm used on implicit data that takes into consider user metadata during content filtering. Matrix factorization is accurate and easy to implement when trained using gradient descend, but it lacks in scalability. The shallow softmax model does not function well on small datasets.

• Keywords: sustavi preporučivanja; suradničko filtriranje; matrična faktorizacija; klasifikacija
• Keywords (english): recommender systems; collaborative filtering; matrix factorization; classification
• Language: croatian
• URN:NBN: urn:nbn:hr:168:628215
• 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: Closed access
• Created on: 2023-10-11 10:54:22