Algoritamske i implementacijske optimizacije pristupa za traženje redeskripcija CLUS-RM

master's thesis

• Title (english): Algorithmic and implementation optimizations of the CLUS-RM redescription search approach
• Author: Ivan Jukić
• Mentor: Matej Mihelčić (mentor)
• Committee member: Matej Mihelčić (predsjednik povjerenstva)
• Committee member: Darko Androić (član povjerenstva)
• Committee member: Mladen Vuković (član povjerenstva)
• Committee member: Dijana Ilišević (član povjerenstva)
• Committee member: Mario Basletić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Mathematics); Zagreb
• Defense date and country: 2022-03-03, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Mathematics

Abstract

U ovom radu opisuje se nekoliko optimizacija algoritma CLUS-RM čija je zadaća pronalazak redeskripcija koristeći višeciljna prediktivna stabla klasteriranja. Prvi dio rada opisuje pretraživanje redeskripcija kao tehniku dubinske analize podataka te navodi definicije i izraze vezane uz područje. Drugi dio rada opisuje algoritam CLUS-RM koji koristi višeciljna prediktivna stabla klasteriranja te opisuje i sam algoritam stvaranja prediktivnih stabala. Treći dio rada prikazuje algoritamsku optimizaciju provjere Jaccard indeksa dvaju pravila u algoritmu CLUS-RM za podatke koji imaju nedostajuću vrijednost i za podatke koji nemaju nedostajuću vrijednost uz analizu složenosti optimizacije. Uz to se pojašnjavaju implementacijske optimizacije provedene nad algoritmom postignute promjenom načina dvosmjerne komunikacije izmedu algoritama CLUS-RM i CLUS. U zadnjem dijelu rada, analiziraju se provedeni eksperimenti i ušteda vremena za obje optimizacije algoritma temeljem tri različita skupa podataka.

Abstract (english)

This thesis describes several approaches to optimize the CLUS-RM algorithm whose task is to find redescriptions using multi-target predictive clustering trees. The first part of the thesis describes the search for redescriptions as a technique of data mining and provides definitions and expressions related to the field. The second part of the thesis describes the CLUS-RM algorithm that uses multi-target predictive clustering trees and describes the algorithm for creating predictive trees. The third part of the thesis presents the algorithmic optimization of the Jaccard index calculation of the two rules in the CLUS-RM algorithm for data with missing values and data without missing values, also including the complexity analyis of the proposed optimization. In addition, the implementation optimizations performed on the algorithm achieved by changing the way of two-way communication between the CLUS-RM and CLUS algorithms are clarified. The last part of the thesis analyses the experiments performed and the time savings gained for each proposed algorithm optimization on three different data sets.

• Keywords: algoritam CLUS-RM; viišeciljna prediktivna stabla klasteriranja; Jaccard indeks
• Keywords (english): CLUS-RM algorithm; multi-target predictive clustering trees; Jaccard index
• Language: croatian
• URN:NBN: urn:nbn:hr:217:418230
• Study programme: Title: Mathematics and Physics Education; specializations in: Mathematics and Physics Education; Course: Mathematics and Physics Education; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra edukacije matematike i fizike (magistar/magistra edukacije matematike i fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-03-28 10:56:19