Analiza molekularnog potpisa tumora uz pomoć strojnog učenja

undergraduate thesis

• Title (english): Analysis of molecular signature of tumors with machine learning
• Author: Josip Jukić
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Nikolina Frid (član povjerenstva)
• Committee member: Marko Đurasević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U uvodnim poglavljima izložena je motivacija problema. Kratko su opisani biološki koncepti čije je poznavanje važno u kasnijim fazama. Glavna tema rada je razvoj generičkog klasifikatora. Naglasak se stavlja na mogućnost određivanja tipova i podtipova
tumorskih stanica. U inkrementalnom postupku razvoja koristi se algoritam stroja potpornih vektora (SVM). Prikazan je način optimizacije parametara SVM-a uz pomoć genetskog algoritma. Opisana je implementirana paralelizacija posla pri učenju.
Izgrađeni klasifikator ispitan je na različitim vrstama skupova podataka. Ukratko su opisane strategije višerazredne klasifikacije. Predstavljen je način procjene vjerojatnosti pripadnosti uzoraka pojedinim razredima klasifikacije u vidu Plattovog skaliranja. Izloženi su rezultati i uspješnost konačnog modela.

Abstract (english)

Motivation for this assignment’s problem is presented in introductory chapters. Biological concepts, key to understanding later work, are briefly described. Main topic in this assignment is development of generic classifier. Emphasis is put on possibility of determining types and subtypes of tumor cells. Algorithm Support Vector Machine is used in incremental building procedure. Optimization of SVM’s parameters with genetic algorithm is described thoroughly. Implementation of parallel computing for work needed for training is also presented. Built classifier is tested on various types of datasets. Strategies of multiclass classification are shortly described. A way of creating a probability of class association approximation is shown by method of Platt scaling. Results and achievements of final model are stated in the end.

• Keywords: klasifikacija; jezgrene funkcije; genetski algoritam; tumorske stanice
• Keywords (english): classification; kernel functions; genetic algorithm; tumor cells
• Language: croatian
• URN:NBN: urn:nbn:hr:168:981723
• 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: 2019-03-13 21:16:09