Domenski ograničen preporučiteljski sustav

master's thesis

• Title (english): Single-domain Recommender System
• Author: Zvonimir Gračak
• Mentor: Ljiljana Brkić (mentor)
• Committee member: Ljiljana Brkić (predsjednik povjerenstva)
• Committee member: Zdravko Galić (član povjerenstva)
• Committee member: Igor Mekterović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-03, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U okviru ovog rada implementirani su collaborative-filtering, content-based i knowledge-based preporučiteljski sustavi na domeni kupnje nekretnina. Implementirani su sustav za prikupljanje dinamičnog skupa podataka o nekretninama s javno dostupnih izvora, preporučiteljski sustavi koji korisnički profil uče na temelju interakcije korisnika sa sustavom te web aplikacija za demonstraciju rada cijelog sustava. Korisniku je kroz korištenje sustava omogućeno ocjenjivanje nekretnina i definiranje vlastitih interesa, na temelju čega preporučiteljski sustavi raspoznaju koji sadržaj bi bio relevantan za korisnika. S obzirom da preporučiteljski sustavi korisnički profil mogu naučiti tek na temelju interakcije korisnika sa sustavom, za savladavanje cold start problema implementirano je stvaranje generičkih preporuka za sve korisnike. Dodatno, za evaluaciju kvalitete preporučiteljskih sustava i poboljšanje budućih preporuka za korisnika, implementiran je i modul za ocjenjivanje preporuka pojedinog preporučiteljskog sustava.

Abstract (english)

This paper describes the development and implementation of collaborative-filtering, content-based and knowledge-based recommender systems on the real estate domain. A dynamic system for gathering data about real estates from publicly available sources is implemented, as well as recommender systems that generate user profiles based on user's interaction with the system. For demonstration purposes, a web application is built. Through using the system, user is presented with the availability to rate real estates and to define own interests. Using that data, recommender systems can determine which content, in this particular case real estates, would be relevant to the user. Based on the fact that in this case recommender systems can learn user profile only after the user interaction with the system, for handling the cold start problem, generating generic recommendations for all users in the system is implemented. Furthermore, for evaluating recommender systems and improving the recommendation quality for users, a module for rating recommendation for each of the recommender system is implemented.

• Keywords: preporučiteljski sustav; collaborative filtering; content-based; knowledge-based; cold start problem; ASP.NET MVC; API; MongoDB
• Keywords (english): recommender system; collaborative filtering; content-based; knowledge-based; cold start problem; ASP.NET MVC; API; MongoDB
• Language: croatian
• URN:NBN: urn:nbn:hr:168:499276
• 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: 2019-03-13 20:32:53