Usporedba metoda grupiranja primjenom programskog jezika Python

undergraduate thesis

• Title (english): Comparison of Clustering Methods Using Python Programming Language
• Author: Tomislav Bjelčić
• Mentor: Goran Delač (mentor)
• Committee member: Goran Delač (predsjednik povjerenstva)
• Committee member: Adrian Satja Kurdija (član povjerenstva)
• Committee member: Lucija Šikić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-09-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Grupiranje je postupak particioniranja skupa neoznačenih podatka na podskupove sa međusobno sličnim podacima. Ovaj rad opisuje osnovne pojmove problema grupiranja i uspoređuje četiri različita algoritama grupiranja na jednostavnim skupovima podataka.

Algoritam K-srednjih vrijednosti je najpoznatiji, najučinkovitiji algoritam grupiranja koji se zasniva na modelu centroida. Zahtjeva poznavanje broja grupa i može grupirati konveksne oblike grupa. Hijerarhijsko aglomerativno grupiranje zasniva se... More na kriteriju spajanja, gradi hijerarhiju grupa spajanjem i nije učinkovit kao K-sredina ali je primjenjiv na svim vrstama podataka. DBSCAN se zasniva na modelu gustoće točaka koje prepoznaje i grupira gusto raspoređene podatke. DBSCAN je brz, poznaje koncept šuma i može grupirati proizvoljne oblike. EM algoritam Gaussovih mješavina nastoji procijeniti parametre modela Gaussovih mješavina koji podatke promatra kao realizacije više Gaussovih razdioba. EM algoritam je brz i može grupirati konveksne grupe raznih oblika.

Opisani algoritmi su pokrenuti, vrednovani i uspoređeni na 6 različita, relativno jednostavna skupa dvodimenzijskih podataka sa realnim značajkama. Less

Abstract (english)

Clustering is a process of partitioning a set of unlabeled data into subsets with similar data. This thesis describes basic concepts regarding cluster analysis and compares four different clustering algorithms on simple data sets.

K-means algorithm is the best known, most efficient clustering algorithm based on the centroid model. K-means requires knowledge of the number of clusters and is able to group convex group shapes. Hierarchical agglomerative clustering is based on linkage criterion,... More builds a hierarchy of clusters by merging them, and is not as efficient as K-means, but is applicable to all types of data. DBSCAN is based on the density model which discovers and clusters dense regions of data. DBSCAN is fast, has a notion of noise and is able to cluster arbitrarily shaped clusters. EM algorithm for Gaussian mixture model aims to estimate parameters of Gaussian mixture model that views the data as realizations of several Gaussian distributions. EM algorithm is fast and is able to cluster convex groups of various shapes.

Described algorithms were executed, evaluated and compared on 6 different, relatively simple sets of two-dimensional data with real number features.
Less

• Keywords: Grupiranje; nenadzirano strojno učenje; mjera udaljenosti
• Keywords (english): Clustering; unsupervised machine learning; distance measure; similarity measure
• Language: croatian
• URN:NBN: urn:nbn:hr:168:676899
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-14 20:05:35