Predstavljanje znanja zasnovano na integraciji ontologija i Bayesovih mreža

doctoral thesis

• Title (english): Knowledge representation based on the integration of ontologies and Bayesian networks
• Author: Marin Prcela
• Mentor: Nikola Bogunović (mentor)
• Committee member: Bojana Dalbelo Bašić (predsjednik povjerenstva)
• Committee member: Dragan Gamberger (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2010-01-28, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

Ontologije temeljene na deskriptivnoj logici imaju semantiku koja je dovoljno izražajna za formalizaciju opisnog znanja (opis domene, terminologija), a u jednom dijelu i proceduralnog znanja (akcije koje se izvršavaju). Baza znanja formalizirana
na taj naĊin osnova je modernih medicinskih sustava za potporu pri odlučivanju koji se u trenutku upotrebe oslanjaju na vanjske izvore podataka o pacijentima - najčešće iz nekih elektronskih registara pacijenata (medicinskih baza podataka). U
slučajevima kada nedostaje dio informacije potreban za donošenje zaključaka, sustav nije u mogućnosti donositi konkretne odluke. U takvim slučajevima, uvođenjem Bayesovih mreža u sustav, mogu se donositi zaključci temeljeni na nepotpunoj informaciji - koji su najvjerojatniji zaključci i koje su najinformativnije akcije koje vode konkretnim zaključcima. U doktorskom se radu kroz primjer dijagnoze srčanog zatajenja predlaže metodologija integracije dvaju već provjerenih oblika predstavljanja znanja: ontologije i Bayesovih mreža. Pritom se analiziraju oblici upotrebe ovako integriranog sustava te se definiraju postupci rasuđivanja u integriranom sustavu. Nadalje, opisuje se formalizam koji omogućuje uključivanje realnih medicinskih podataka promatranog pacijenta u ovako oblikovanoj bazi znanja. Na kraju se daje uvid u korisničko sučelje sustava te se analiziraju neke funkcionalne značajke sustava kao što su vrijeme izvođenja i pogreška učinjena pri rasuđivanju.

Abstract (english)

Ontologies based on description logic have expressive enough semantics to cover the
descriptive part of knowledge base (domain description, terminology), and also a large
part of procedural knowledge in the knowledge base (actions to be taken). Such encoded
knowledge is the basis of modern medical decision support systems, which at the time of
use also have rely on external patient data – typically from some electronic health record.
In cases when some amount of information is missing the system often has to refrain to
draw any concrete conclusion. In such cases by introducing Bayesian networks into the
system conclusions can be made based on uncertain knowledge: what are the most
probable conclusions and what are the most informative actions that possibly lead to
concrete conclusions. Through the example of diagnosis of heart failure disease this
dissertation demonstrates methodology for integration of two well known forms of
knowledge representation: ontologies and Bayesian networks. Features of such integrated
system are analyzed and reasoning procedures are defined. Furthermore, a formalism that
includes data of observed patients into reasoning process is described. Finally, user
interface of the example decision support system is described and some functional
properties of the system are analyzed such as duration of reasoning and error made in
reasoning.

• Keywords: predstavljanje znanja; ontologije; Bayesove mreže; deskriptivna logika; integracija; informacijska dobit; sustavi za potporu pri odlučivanju; ekspertni sustavi
• Keywords (english): knowledge representation; ontologies; Bayesian networks; description logics; integration; information gain; decision support system; expert system
• Language: croatian
• URN:NBN: urn:nbn:hr:168:192256
• Promotion: 2010
• Study programme: Title: Computer Science; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti (Doktor znanosti)
• Type of resource: Text
• Extent: 150 str. : graf. prikazi
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-03-19 10:50:08