Implementacija i evaluacija preporučiteljskih sustava temeljenih na neuronskim mrežama

master's thesis

• Title (english): Implementation and Evaluation of Recommender Systems based on Neural Networks
• Author: Lovro Vidović
• Mentor: Krešimir Pripužić (mentor)
• Committee member: Krešimir Pripužić (predsjednik povjerenstva)
• Committee member: Dragan Jevtić (član povjerenstva)
• Committee member: Marin Vuković (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-09, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U okviru rada bavimo se usporedbom klasičnih preporučiteljskih sustava temeljenih na suradnji korisnika i preporučiteljskih sustava temeljenih na neuronskim mrežama i drugim aspektima dubokog učenja. Opisani su teoretski aspekti metoda preporučivanja i duboke neuronske mreže. Koristeći Python i razvojni okvir Tensorflow implementirana su 4 preporučiteljska sustava: preporučiteljski sustav temeljen na suradnji korisnika, preporučiteljski sustav temeljen na modelu latentnih faktora, preporučiteljski sustav temeljen na dubokim neuronskim mrežama i preporučiteljski sustav temeljen na ansamblu modela latentnih faktora i dubokih neuronskih mreža. Preporučiteljski sustavi su evaluirani nad javno dostupnom podatkovnom skupu MovieLens. Rezultati evaluacije su grafički prikazani. Analizirali smo mogućnost raspodijeljenog izvođenja preporučiteljskih algoritama u računalnom grozdu.

Abstract (english)

In this thesis we are focusing on comparison between user-based collaborative filtering and neural network based recommender systems with usage of numerous deep learning aspects. Theoretical aspects of the recommender systems and deep neural network are described. Using Python and TensorFlow framework, four recommender systems were implemented: user-based collaborative filtering, latent factor model recommender system, deep neural network based recommender system and an ensemble of latent factor model and deep neural network model. Recommender systems were evaluated over the publicly available MovieLens dataset. The evaluation results are shown graphically. Furthermore, we have analyzed distributed performance of recommender systems on a computer cluster.

• Keywords: preporučiteljski sustavi; suradno filtriranje; duboke neuronske mreže; model latentnih faktora; faktorizacija matrice; raspodijeljena obrada
• Keywords (english): recommender systems; collaborative filtering; deep neural networks; latent factor model; matrix factorization; distributed computing
• Language: croatian
• URN:NBN: urn:nbn:hr:168:250102
• 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: 2019-03-14 01:55:48