magistrsko delo
Katja Nemec (Author), Mario Gorenjak (Mentor), Uroš Potočnik (Co-mentor)

Abstract

Uvod: V našem magistrskem delu smo želeli ugotoviti učinkovitost metod strojnega učenja pri napovedi odziva bolnikov s Chronovo boleznijo na biološko zdravilo adalimumab. Metode: Raziskava je vključevala 88 vzorcev, ki so bili analizirani glede na genetske, klinične in mešane podatke v različnih tednih zdravljenja. Uporabljene metode, kot so naključni gozdovi (RF), podporni vektorji (SVM) in nevronske mreže (NNET), so bile evalvirane z uporabo različnih metrik natančnosti, občutljivosti in Youdenovega indeksa. Rezultati: Rezultati kažejo, da je metoda RF najboljša na mešanih podatkih, SVM izstopa pri kliničnih, medtem ko NNET in RF dosegata najboljše rezultate na genetskih podatkih v različnih obdobjih zdravljenja. Uporaba metode "bagging" je izboljšala natančnost, še posebej pri RF. Kljub temu se zahteva previdnost pri interpretaciji zaradi omejene velikosti vzorca. Razprava: Naša analiza poudarja potrebo po preudarnem izboru metode, odvisnem od specifičnih značilnosti podatkov in ciljev analize. Sklep: Naše ugotovitve na podlagi analize predstavljajo osnovo za nadaljnje raziskave v smeri izboljšanja natančnosti modelov napovedi zdravljenja.

Keywords

Crohnova bolezen;bioinformatika;napovedni modeli;strojno učenje;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UM - University of Maribor
Publisher: [K. Nemec]
UDC: 004.43:615.32:616.34-002(043.2)
COBISS: 190297859 Link will open in a new window
Views: 219
Downloads: 9
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Comparison of support vectors, random forests and neural networks for predicting the response to adalimumab treatment in slovenian patients with crohn´s disease
Secondary abstract: Introduction: In our master's thesis, we aimed to assess the effectiveness of machine learning methods in predicting the response of patients with Crohn's disease to the biological drug adalimumab. Methods: The study involved 88 samples, analyzed based on genetic, clinical, and combined data over various treatment weeks. Employed methods, such as Random Forest (RF), Support Vector Machine (SVM), and Neural Network (NNET), were evaluated using diverse accuracy metrics, sensitivity, and the Youden Index. Results: Findings indicate RF as optimal for mixed data, SVM excelling in clinical data, while NNET and RF performed best on genetic data across different treatment periods. The use of "bagging" improved accuracy, particularly with RF. However, caution is warranted in interpretation due to the limited sample size. Discussion: Our analysis underscores the need for a judicious method selection, contingent on specific data characteristics and analysis goals. Conclusion: Our insights, derived from this analysis, serve as a foundation for further research aimed at enhancing the accuracy of treatment prediction models.
Secondary keywords: Crohn`s disease;bioinformatics;prediction models;machine learning;Crohn Disease;Neural Pathways;Patients;Machine Learning;Crohnova bolezen;Nevronske poti;Bolniki;Strojno učenje;
Type (COBISS): Master's thesis
Thesis comment: Univ. v Mariboru, Fak. za zdravstvene vede
Pages: XVII, 81 str.
ID: 22700587