Vzporedni homogeni ansambli napovednih modelov: naključni gozd : delo diplomskega seminarja

Lindič, Rok (author), Todorovski, Ljupčo (mentor)

V strojnem učenju uporabljamo ansamble napovednih modelov, da zmanjšamo napako, ki bi se pojavila ob uporabi samo enega modela. Pogosto uporabljena ansambelska metoda je naključni gozd. V diplomskem delu bomo opisali, kako naključni gozd deluje, kaj so njegove prednosti v primerjavi z odločitvenimi drevesi, na koncu pa bomo preučili vpliv raznolikosti osnovnih modelov v ansamblih na njihovo napovedno napako.

• Language: Slovenian
• Keywords: strojno učenje, odločitvena drevesa, ansambli, homogeni ansambli, naključni gozdovi, dekompozicija napovedne napake
• Work type: Final seminar paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FMF - Faculty of Mathematics and Physics
• Year: 2025
• PID: 20.500.12556/RUL-170594
• UDC: 004.8
• COBISS.SI-ID: 242139907
• Publication date in RUL: 10.07.2025

Secondary language

• Language: English
• Title: Homogenous parallel ensemble methods: random forest

Abstract

In machine learning, ensembles of prediction models are used in order to reduce the error that would occur if only one model was used. A commonly used ensemble method is the random forest. The thesis will describe how the random forest functions and what its advantages are compared to decision trees. The end of the thesis will focus on the impact that the diversity of the underlying models in the ensembles has on their prediction error.

• Keywords: machine learning, decision trees, ensembles, homogenous ensembles, random forest, bias-variance decomposition