Vpliv podobnosti na uspešnost klasifikacije evolucijskih odločitvenih dreves

magistrsko delo

Leon Bošnjak (Author), Vili Podgorelec (Mentor)

Abstract

Magistrska naloga obravnava proces gradnje klasifikacijskih odločitvenih dreves z genetskimi algoritmi, v sklopu katerega se osredotoča na ocenjevanje uspešnosti zgrajenih dreves ter hitrosti oziroma učinkovitosti algoritma. Standardni način evolucijske gradnje odločitvenih dreves predvideva uporabo naključne selekcije dveh primerkov za križanje dreves, kar lahko povzroči prehitro konvergenco k lokalno optimalni rešitvi. Z namenom ohranjanja raznolikosti populacije tekom evolucije je bilo implementiranih pet pristopov vrednotenja podobnosti med drevesi, ki so bili uporabljeni v okviru selekcije primerkov za križanje. Pristopi križanja med seboj različnih in podobnih dreves so bili primerjani s standardnim načinom brez upoštevanja podobnosti na enaindvajsetih različnih podatkovnih množicah z namenom ugotavljanja vpliva podobnosti na uspešnost in učinkovitost algoritma.

Keywords

odločitvena drevesa;genetski algoritmi;klasifikacija;podobnosti;

Data

Language:	Slovenian
Year of publishing:	2014
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[L. Bošnjak]
UDC:	659.21:316.773.3(043.2)
COBISS:	17980694
Views:	1551
Downloads:	214
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	The impact of similarity on the classification performance of evolutionary decision trees
Secondary abstract:	The master's thesis deals with the process of building classification decision trees with genetic algorithms, focusing on the assessment of performance of constructed trees, as well as the speed and efficiency of the algorithm. The standard evolutionary method of building decision trees assumes the use of random selection of two trees for crossover, which can lead to premature convergence to a local, often sub-optimal solution. In order to maintain the diversity of the population over the course of evolution, five different approaches to evaluate the similarity between trees were implemented. The approaches of both similar and diverse tree crossover were compared to the standard approach on twenty-one different data sets to determine the impact of similarity on the effectiveness and efficiency of the algorithm.
Secondary keywords:	decision trees;genetic algorithms;classification;similarity;
URN:	URN:SI:UM:
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko
Pages:	VIII, 104 str.
ID:	8729522