Potraga za pulsarima primjenom strojnog učenja

master's thesis

• Title (english): Search for pulsars using machine learning
• Author: Marin Bilić
• Mentor: Goranka Bilalbegović (mentor)
• Committee member: Goranka Bilalbegović (predsjednik povjerenstva)
• Committee member: Krešimir Kumerički (član povjerenstva)
• Committee member: Mihael Makek (član povjerenstva)
• Committee member: Ivica Smolić (član povjerenstva)
• Committee member: Dalibor Paar (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2020-09-22, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Pulsari su brzo rotirajuće neutronske zvijezde. Ti vrlo zanimljivi objekti fasciniraju astronome još od trenutka kad su prvi put pronađeni. S vremenom su teleskopi postajali sve bolji te se tako i broj otkrivenih pulsara povećavao. Ručno analiziranje podataka dobivenih pomoću teleskopa u potrazi za pulsarima je vrlo naporan i monoton proces koji zahtjeva puno vremena. Međutim, razvojem strojnog učenja je ovaj zadatak postao rješiv pomoću računala. U ovom diplomskom radu analizirani su pripremljeni podaci s kandidatima za pulsare. Njihovom obradom pomoću tri različita algoritma strojnog učenja razvijeni su modeli kategorizacije koji daju veliku preciznost. Korišteni su algoritmi k-najbližih susjeda, potpornih vektora i stabla odlučivanja. Ti algoritmi su analizirani i testirani za više vrijednosti parametara kako bi se dobila najbolja rješenja. Uspoređeni su rezultati primijenjenih algoritama. U metodičkom dijelu je dan osvrt na teme iz astronomije i pulsara u osnovnim i srednjim školama, te je razrađen mogući uvod u strojno učenje za srednjoškolsku nastavu.

Abstract (english)

Pulsars are neutron stars with fast rotation speeds. These very interesting objects fascinate astronomers from the moment they were discovered. With time, telescopes are becoming better, therefore the number of discovered pulsars is growing. Analyzing the telescope data by hand is a very tedious and monotone process. However, with development of machine learning this task became easily solvable by using a computer. In this diploma thesis we analyze the prepared data with pulsar candidates. Three different machine learning algorithms are applied and they produced prediction models with high accuracy. Algorithms applied are: k-nearest neighbors, support vectors and decision tree. All algorithms are analyzed and tested for several values of parameters in order to produce the best solution. Results of applied algorithms are compared. In the methodical section, a summary of topics in astronomy and pulsars for middle and high school students is given. In addition, a possible introduction to machine learning for high school students is formulated.

• Keywords: pulsari; strojno učenje; algoritam k-najbližih susjeda; algoritam potpornih vektora; algoritam stabla odlučivanja
• Keywords (english): pulsars; machine learning; K-nearest neighbors algorithm; support vectors algorithm; decision tree algorithm
• Language: croatian
• URN:NBN: urn:nbn:hr:217:327347
• Study programme: Title: Physics and Computer Science; specializations in: Education; Course: Education; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra fizike i informatike (magistar/magistra fizike i informatike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2021-01-19 09:56:52