Metode strojnog učenja u određivanju tržišne vrijednosti igrača u nogometu

master's thesis

• Title (english): Machine learning methods in determining the market value of football players
• Author: Ivan Čulin
• Mentor: Nikola Sandrić (mentor)
• Committee member: Nikola Sandrić (predsjednik povjerenstva)
• Committee member: Igor Pažanin (član povjerenstva)
• Committee member: Sonja Štimac (član povjerenstva)
• Committee member: Miljenko Huzak (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Mathematics); Zagreb
• Defense date and country: 2024-03-01, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Mathematics

Abstract

Ovaj diplomski rad bavi se predviđanjem tržišne vrijednosti nogometaša koristeći metode statističkog učenja. Na početku rada podsjetili smo se linearnog regresijskog modela i najpoznatije metode njegove prilagodbe - metode najmanjih kvadrata. U nastavku definiramo osnovne pojmove statističkog učenja, te uvodimo metode odabira prediktora u linearnim modelima i metode regularizacije (ridge i lasso regresija). Svaku od metoda smo primijenili na podatke o nogometašima. Rezultati ukazuju na postojanje određene linearne veze između prediktora i varijable odziva uz očite znakove potrebe za povećanjem kompleksnosti modela. Iako sve metode dovode do približno sličnih rezultata, kao najpouzdanija metoda pokazala se ridge regresija s obzirom da je njena pripadna testna greška najmanja.

Abstract (english)

The main goal of this thesis is to predict the market value of football players using methods of statistical learning. At the beginning of the paper, we introduce the linear regression model and the least squares method. Furthermore, we defined the basic concepts of statistical learning and introduced methods for predictor selection in linear models and regularization methods (ridge and lasso regression). Each of these methods was applied to football player data. The results indicate the existence of a certain linear relationship between predictors and the response variable with obvious signs of the need to increase the model's complexity. Although all methods lead to approximately similar results, ridge regression proved to be the most reliable method since its corresponding test error is the lowest.

• Keywords: metode statističkog učenja; linearni regresijski model; metoda najmanjih kvadrata; ridge regresija; lasso regresija
• Language: croatian
• URN:NBN: urn:nbn:hr:217:972343
• Study programme: Title: Mathematical Statistics; Study programme type: university; Study level: graduate; Academic / professional title: sveučilišni magistar matematike (sveučilišni magistar matematike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Embargoed access; Embargo expiration date: 2025-03-13
• Created on: 2024-03-13 08:11:07