Traženje sličnosti u podacima kontinuiranog praćenja glukoze

undergraduate thesis

• Title (english): Similarity Search in Continuous Glucose Monitoring Data
• Author: Krunoslav Jurčić
• Mentor: Ratko Magjarević (mentor)
• Committee member: Ratko Magjarević (predsjednik povjerenstva)
• Committee member: Sara Žulj (član povjerenstva)
• Committee member: Goran Šeketa (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-09-02, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

Temeljna ideja ovog rada bila je istražiti metode traženja sličnosti vremenskih nizova, s posebnim naglaskom na fiziološke i biološke vremenske nizove, te onda procijeniti koje metode bi trebale donositi najpreciznije rezultate. Nakon odabira, metode su implementirane u program za usporedbu odsječaka zapisa o kontinuiranom praćenju razine glukoze u krvi sa zadanim uzorkom u trajanju od sat vremena, sa vremenom uzorkovanja 5 minuta. Za implementaciju je korišten programski jezik Matlab. Tokom razvijanja programa te nakon završne implementacije tri metode traženja sličnosti su se ispostavile najpreciznijima pri radu s ovom vrstom podataka: metoda Euklidske udaljenosti, wavelet dekompozicija i dinamičko vremensko savijanje (DTW). Za svaku metodu su izračunate sličnosti za svaki odsječak zapisa. Rezultati su pokazali da su sve tri metode dovoljno precizne i vjerodostojne da bi se mogle koristiti pri traženju sličnosti među signalima razine glukoze u krvi, što bi se moglo pokazati veoma korisnim pri predikciji promjene razine glukoze u krvi u budućnosti.

Abstract (english)

The main idea of this work was to explore the methods of similarity search using time series, with the main focus being on physiological and biological time series, and choose the methods best suitable for these types of time series. After that, the methods are implemented to a programme for comparison between parts of a continuous glucose monitoring time series and a sample with a full duration of one hour, time sampling being 5 minutes. Programming language used for the implementation of this programme was Matlab. During the development of the programme and after the final implementation three methods of similarity search proved most precise with this type of dana: Euclidean distance, wavelet decomposition and dynamic time warping. Similarity search was made using every one of those methods. The results showed that all three methods provided valuable and useful information, and that all of them could be used when searching for similarity between glucose signals, which might prove very useful for future prediction of change of glucose level.

• Keywords: praćenje glukoze; traženje sličnosti; Euklidska udaljenost; wavelet dekompozicija; DTW
• Keywords (english): glucose monitoring; similarity search; Euclidean distance; wavelet decomposition; dynamic time warping
• Language: croatian
• URN:NBN: urn:nbn:hr:168:918507
• Study programme: Title: Electrical Engineering and Information Technology; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka elektrotehnike i informacijske tehnologije (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka elektrotehnike i informacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-01-07 18:49:11