undergraduate thesis
K-nearaest neighbors algorithm and its improved versions for pattern recognition

Marin Radoš (2015)
Sveučilište Josipa Jurja Strossmayera u Osijeku
Fakultet elektrotehnike, računarstva i informacijskih tehnologija Osijek
Zavod za programsko inženjerstvo
Katedra za programske jezike i sustave
Metadata
TitleAlgoritam k-najbližih susjeda i njegove unaprijeđene inačice za prepoznavanje uzoraka
AuthorMarin Radoš
Mentor(s)Goran Martinović (thesis advisor)
Abstract
U ovom radu prikazan je problem klasifikacije uzoraka pomoću algoritma k-najbližih susjeda, te unaprjeđenje tog algoritma raznim načinima. Algoritam k-najbližih susjeda uspoređen je s njegovim unaprijeđenim inačicama na temelju eksperimentalnih rezultata dobivenih mjerenjem točnosti klasifikacije u različitim uvjetima, s različitim skupovima podataka i brojem susjeda k. Određeni su najveći nedostatci standardnog algoritma k-najbližih susjeda kroz funkciju udaljenosti, dodjeljivanje jednake važnosti svim atributima, te raspon veličina atributa. Ti nedostatci su riješeni u obliku unaprjeđenja algoritma drugačijom funkcijom udaljenosti, u ovom radu Manhattan, dodavanjem težine važnijim atributima, te normalizacijom skupa podataka. Kroz sva tri unaprjeđenja vidljivo je povećanje točnosti klasifikacije kroz dobivene eksperimentalne rezultate.
Keywordseuclidean distance classification nearest neighbors normalization
Parallel title (English)K-nearaest neighbors algorithm and its improved versions for pattern recognition
Committee MembersGoran Martinović (committee chairperson)
Hrvoje Glavaš (committee member)
Tomislav Keser (committee member)
GranterSveučilište Josipa Jurja Strossmayera u Osijeku
Fakultet elektrotehnike, računarstva i informacijskih tehnologija Osijek
Lower level organizational unitsZavod za programsko inženjerstvo
Katedra za programske jezike i sustave
PlaceOsijek
StateCroatia
Scientific field, discipline, subdisciplineTECHNICAL SCIENCES
Computing
Program Engineering
Study programme typeuniversity
Study levelundergraduate
Study programmeUndergraduate University Study Programme in Computer Engineering
Academic title abbreviationuniv.bacc.ing.comp.
Genreundergraduate thesis
Language Croatian
Defense date2015-09-24
Parallel abstract (English)
In this paper, the problem of classification was shown using the k-nearest neighbors algorithm and its improvements through various ways. K-nearest neighbors algorithm was compared with its improved variants based on the experimental results obtained by measuring classification accuracy in different conditions, with different data sets and the number of neighbors. Biggest disadvantages of the standard k-nearest neighbors algorithm were defined through a distance function, assigning equal importance to all the attributes, and size range of the attributes. These shortcomings are addressed in the form of improving the algorithm with different distance function, in this paper, Manhattan, adding weight to important attributes, and normalization of the data set. Through all three improvements, an increase in the classification accuracy is shown through the experimental results
Parallel keywords (Croatian)euklidska udaljenost klasifikacija najbliži susjedi normalizacija
Resource typetext
Access conditionOpen access
Terms of usehttp://rightsstatements.org/vocab/InC/1.0/
URN:NBNhttps://urn.nsk.hr/urn:nbn:hr:200:212400
CommitterAnka Ovničević