Metoda samonadziranog učenja za sustave preporuke temeljena na detekciji zajednica u grafovima

undergraduate thesis

• Title (english): Self-supervised learning method for recommender systems based on community detection in graphs
• Author: Dora Pavelić
• Mentor: Mario Brčić (mentor)
• Committee member: Mario Brčić (predsjednik povjerenstva)
• Committee member: Agneza Krajna (član povjerenstva)
• Committee member: Juraj Dončević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-09-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Poznati izazov u sustavima preporuke jest tzv. problem hladnog starta koji nastaje u ranoj fazi postprodukcije kada još nemamo zabilježene korisničke preference u obliku recenzija na temelju kojih bi sustav mogao generirati preporuke. U okviru rada, predloženo je rješenje koje koristi metodu samonadziranog učenja kojim se neuronska mreža uči reprezentacijama korisnika i sadržaja na temelju kojih se uz međusobnu usporedbu dolazi do najpoželjnije preporuke. Kako bi se dokazala uspješnost ove metode, dizajniran je sustav preporuke znanstvenih članaka. Rezultati ovog pristupa su zadovoljavajući. Kako bi se u budućnosti postigli još bolji rezultati, naučeni model trebao bi se iskoristiti u sklopu prijenosnog učenja (eng. transfer learning) uz prave korisničke recenzije.

Abstract (english)

A well-known challenge in recommender systems is the so-called cold start problem that arises in the early stages of post-production when we do not yet have recorded user preferences in the form of reviews, based on which the system could generate recommendations. This final paper suggest a solution based on self-supervised learning in which the neural network learns the representations of users and the content on the basis of which they are compared to the most desirable recommendations. As part of proving proposed method, a system of recommending scientific articles has been developed. The results of this approach are satisfactory; however, in order to achieve even better results, the developed trained model should be used in transfer learning with real user reviews.

• Keywords: Sustavi preporuke; Hladni start; Samonadzirano učenje; Detekcija zajednica u grafovima; Obrada prirodnog jezika
• Keywords (english): Recommender systems; Cold start; Self-supervised learning; Community detection; Natural language processing
• Language: croatian
• URN:NBN: urn:nbn:hr:168:971823
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-15 10:59:32