Objašnjivost klasifikacijskih modela u zdravstvenoj domeni

undergraduate thesis

• Title (english): Explainability of classification models in the health domain
• Author: Doria Herceg
• Mentor: Mihaela Vranić (mentor)
• Committee member: Mihaela Vranić (predsjednik povjerenstva)
• Committee member: Damir Pintar (član povjerenstva)
• Committee member: Ana Šarčević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-08-31, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Na početku rada opisana je motivacija za rad. Da bi modeli strojnog učenja bili primjenjivi u zdravstvenoj domeni, potrebno ih je analizirati te proučiti kako rade jer bez objašnjivosti modela nije moguće vjerovati rezultatima. Pogotovo kada je područje visokorizično poput zdravstva.
U nastavku objašnjena je teorija dva modela iz područja nadziranog učenja: Naivni Bayesov klasifikator te ID3 stablo odlučivanja. Zatim su predstavljene metode analize modela kojima će se evaluirati klasifikatori.
Opisani su skupovi podataka koji su korišteni za učenje i testiranje modela te su predstavljeni rezultati obrade. Iz rezultata i diskusije vidljiva je važnost dobro odabranih skupova podataka za pojedini model, odnosno da će model dobro raditi samo ako su podaci na kojima vrši obradu odgovarajućeg oblika.
Oba su modela zadovoljivo dobrih performansi u slučajevima kada su im na obradu dani podaci oblikovani u skladu sa zahtjevima tog modela.

Abstract (english)

The motivation for the work is described at the beginning of the thesis. In order for machine learning models to be applicable in the health domain, it is necessary to analyze them and study how they work, because without the explainability of the model, it is not possible to trust the results. Especially, when the area of study is as risky as healthcare.
In the following, the theory of two models from the field of supervised learning is explained: Naive Bayes classifier and ID3 decision tree. Then, the model analysis methods used to evaluate the classifiers are presented.
The datasets used for learning and testing the model are described and the results of processing are presented. The results and discussion show the importance of well-chosen data sets for a particular model, that is, that the model will work well only if the data is of the appropriate form.
Both models provide good performance in cases where the data given to them for processing is formatted in accordance with the requirements of that model.

• Keywords: strojno učenje; zdravstvena domena; klasifikatori; Naivni Bayesov klasifikator; ID3 stablo odlučivanja; Python; Java
• Keywords (english): machine learning; health domain; classifiers; Naive Bayes classifier; ID3 decision tree; Python; Java
• Language: croatian
• URN:NBN: urn:nbn:hr:168:248488
• 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: 2023-10-11 09:01:24