Razvoj in implementacija klasifikatorja v okolju Weka

diplomsko delo

Lovro Mažgon (Author), Vili Podgorelec (Mentor)

Abstract

Diplomsko delo se nanaša na področje odkrivanja znanja iz podatkov, še natančneje pa opisuje klasifikacijo, priznane klasifikatorje ter mere za določanje kakovosti klasifikacije. V delu smo predstavili razvoj lastnega algoritma za klasifikacijo, ki s pomočjo izračunov oddaljenosti od učnih primerkov določi razred neznanemu primerku. Podali smo matematično definicijo algoritma ter opis implementacije v okolju Weka, s pomočjo katerega smo preizkusili uspešnost klasifikacije na dveh realnih medicinskih primerih. Dobljeni rezultati nakazujejo, da razviti algoritem razmeroma uspešno klasificira primerke in se lahko primerja s priznanimi klasifikatorji.

Keywords

strojno učenje;podatkovno rudarjenje;klasifikacija;Weka;

Data

Language:	Slovenian
Year of publishing:	2013
Typology:	2.11 - Undergraduate Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[L. Mažgon]
UDC:	004.6(043.2)
COBISS:	17470486
Views:	1961
Downloads:	224
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	DESIGN AND IMPLEMENTATION OF A CLASSIFIER IN WEKA
Secondary abstract:	The thesis addresses the area of knowledge discovery from data, and even more specifically, it describesthe classification, the recognized classifiers and the metrics for evaluating classifier performance. We have presented the development of our own algorithm for classification which determines an unknown sample’s class with the help of distance calculations from the learning samples. The mathematical definition of the algorithm has been provided, as well as the description of its implementation in the Weka environment, through which we have tested the performance of the classifier on two real medical cases. The results indicate that the algorithm classifies samples relatively successfully and it can be compared with recognized classifiers.
Secondary keywords:	machine learning;data mining;classification;Weka;
URN:	URN:SI:UM:
Type (COBISS):	Bachelor thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Informatika in tehnologije komuniciranja
Pages:	VII, 52 str.
ID:	8727690