Sustavi preporučivanja temeljeni na korisničkim sjednicama

master's thesis

• Title (english): Session-based Recommender Systems
• Author: Matej Luburić
• Mentor: Alan Jović (mentor)
• Committee member: Alan Jović (predsjednik povjerenstva)
• Committee member: Vlado Sruk (član povjerenstva)
• Committee member: Marin Golub (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-09-21, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Sustavi preporučivanja temeljeni na korisničkim sjednicama (engl. Session-based recommender systems–SBRS) pružaju preporuke isključivo na temelju interakcija korisnika u trenutnoj sjednici i ne zahtijevaju postojanje korisničkih profila ili njihovih cjelokupnih povijesnih aktivnosti. Cilj takvih sustava je da na temelju niza kliknutih artikala u sjednici preporučuje artikle koji su najizgledniji da će na jednog od njih kliknuti korisnik u sljedećem koraku. U ovom diplomskom radu opisani su i primjenjeni modeli koji daju najbolje rezultate na navedenom problemu u području e-trgovine. Model BERT4Rec, originalno objavljen za sustave preporučivanja koji koriste sve prošle sjednice istog korisnika (engl. session-aware recommendation systems), detaljno je objašnjen i prilagođen navedenom problemu. Svi modeli evaluirani su uobičajenim tehnikama za sustave preporučivanja i međusobno uspoređeni. Na kraju rada navedeni su potencijalni smjerovima razvoja ovog područja.

Abstract (english)

Session-based recommender systems (SBRS) provide recommendations based solely on user interactions in current session and do not require the existence of user profiles or their entire historical activities. Based on a series of clicked items in a session, the goal of such systems is to recommend items that are most likely to be clicked on by the user in the next step. In this master thesis, the models that give the best results for the mentioned problem in the field of e-commerce are described and applied. The BERT4Rec model, originally published for session-aware recommender systems, is explained in detail and adapted to the given problem. All models were evaluated using common techniques for evaluating recommender systems and compared with each other. The potential directions of the development of this area are presented at the end of the thesis.

• Keywords: sustavi preporučivanja; sjednica; klikovi; e-trgovina; GRU4Rec; KNN; BERT4Rec
• Keywords (english): recommender systems; session; click; e-commerce; GRU4Rec; KNN; BERT4Rec
• Language: croatian
• URN:NBN: urn:nbn:hr:168:600167
• 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: 2023-01-14 09:19:03