Ispitivanje utjecaja hiperparametara dubokih neuronskih mreža na učinkovitost klasifikacije poremećaja iz 12-kanalnog elektrokardiograma

undergraduate thesis

• Title (english): Evaluating the Influence of Deep Neural Network Hyperparameters on Disorder Classification Efficiency from 12-Channel Electrocardiogram
• Author: Marin Matjanec
• Mentor: Alan Jović (mentor)
• Committee member: Alan Jović (predsjednik povjerenstva)
• Committee member: Igor Stančin (član povjerenstva)
• Committee member: Josip Renić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-05, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Kardiovaskularne bolesti pripadaju najopasnijim bolestima s kojima se ljudi susreću. Elektrokardiogram najbitniji je nalaz kod svih kardioloških pregleda. Analiza EKG-a problem je koji se može primijeniti u računarstvu koristeći umjetnu inteligenciju, a kao najbolja metoda pokazale su se duboke neuronske mreže. U ovom radu provedeno je istraživanje nekih od najbitnijih hiperparametara dubokih neuronskih mreža, te njihov utjecaj na uspješnost klasifikacije oboljenja iz zapisa EKG-a. Kroz istraživanje točnost početnog modela na skupu za validaciju poboljšana je za 6,3%, a na skupu za testiranje za 5%. Osim analize, kao rezultat ovog rada napravljena je i aplikacija koja korisnicima nudi učitavanje zapisa, njihovo učenje ili testiranje te ispis rezultata.

Abstract (english)

Cardiovascular diseases belong to the most dangerous diseases that people encounter. The electrocardiogram is the most important finding in all cardiac examinations. ECG analysis is a problem that can be applied in computing using artificial intelligence, and deep neural networks have proven to be the best method. In this paper, a research of the most important hyperparameters of deep neural networks was conducted, and their impact on the success of the classification of diseases from ECG records. Through research, the accuracy of the initial model on the validation set was improved by 6.3% and on the test set by 5%. In addition to the analysis, as a result of this paper, an application was created that offers users to load records, learn or test them and print the results.

• Keywords: elektrokardiogram; srce; duboka neuronska mreža; hiperparametar; konvolucijski sloj; sloj s dugim kratokročnim pamćenjem
• Keywords (english): electrocardiogram; heart; deep neural network; hyperparameter; convolutional layer; long short term memory layer
• Language: croatian
• URN:NBN: urn:nbn:hr:168:697375
• 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 21:09:10