PENINGKATAN KINERJA KLASIFIKASI ARITMIA PADA SINYAL FETAL ELEKTROKARDIOGRAM BERBASIS DEEP LEARNING

ERIZKA, ROSA JULIA and Nurmaini, Siti (2025) PENINGKATAN KINERJA KLASIFIKASI ARITMIA PADA SINYAL FETAL ELEKTROKARDIOGRAM BERBASIS DEEP LEARNING. Undergraduate thesis, Sriwijaya University.

Preview	Image RAMA_56201_09011282126105_cover.jpg - Cover Image Available under License Creative Commons Public Domain Dedication. Download (487kB) | Preview
	Text RAMA_56201_09011282126105.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (3MB) | Request a copy
	Text RAMA_56201_09011282126105_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (8MB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (1MB)
	Text RAMA_56201_09011282126105_0002085908_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (639kB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (428kB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (263kB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (238kB) | Request a copy
	Text RAMA_56201_09011282126105_0002085908_07_lamp.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (802kB) | Request a copy

Abstract

This study aims to improve the performance of arrhythmia classification in deep learning models using the Teaching Learning Based Optimization (TLO) algorithm for feature optimization. The features used are extracted from the time domain of RR intervals obtained during the preprocessing stage. The NIFEA-DB and NIFECG datasets are utilized in this study, with R-peaks extracted using the Heartpy library in Python and from the annotation files provided in the dataset. The results of the RR interval are used to calculate time domain features, namely mean_RR, std_RR, nn_50, rms_RR, pnn_50, cv_RR, min_RR, and max_RR. The classification models used are DNN, CNN, LSTM, and BiLSTM, which are compared with those using TLO. The results show that TLO provides good consistency with the validation data and test data. The DNN+TLO, CNN+TLO, and BiLSTM+TLO models produce the same metric results, namely an accuracy of 95.6%, recall and specificity of 97.7%, and precision of 75%, indicating that TLO can help improve model performance. The LSTM+TLO model showed a decline in performance due to overfitting, particularly in the test data. The deep learning models BiLSTM+TLO and DNN+TLO exhibited more stable accuracy and loss curves, and the confusion matrix for both the validation and test (unseen) data accurately predicted each class. Therefore, the BiLSTM+TLO and DNN+TLO models are the best models for classifying arrhythmias in the fetal ECG dataset.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Fetal ECG, NIFEA-DB, NIFECG, arrhythmia, deep learning, TLO
Subjects:	Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Divisions:	09-Faculty of Computer Science > 56201-Computer Systems (S1)
Depositing User:	Rosa Julia Erizka
Date Deposited:	15 Jul 2025 07:09
Last Modified:	15 Jul 2025 07:09
URI:	http://repository.unsri.ac.id/id/eprint/178585

Actions (login required)

View Item