ECG ARRHYTHMIA TIME SERIES CLASSIFICATION USING 1D-CONVOLUTION LSTM NEURAL NETWORKS

Mohammad Qasem Alqaisi, Yousra (2021) ECG ARRHYTHMIA TIME SERIES CLASSIFICATION USING 1D-CONVOLUTION LSTM NEURAL NETWORKS. Magister thesis, Universitas Brawijaya.

Abstract

An electrocardiogram (ECG) can be dependably used as a measuring device to monitor cardiovascular function. The abnormal heartbeat appears in the ECG pattern and these abnormal signals are called arrhythmias. A faster and more accurate result can be reached by classifying and automatically detecting arrhythmia signals. Several machine learning approaches have been applied to enhance the accuracy of results and increase the speed and robustness of models. This research proposes a method based on Timeseries Classification using deep Convolutional -LSTM neural networks and Discrete Wavelet Transform to classify beats in three experiments, the first one is to classify 4 different types of Arrhythmia in the MIT-BIH Database. The second one for enhancement the first experimental results. The third one is for classifying the whole MIT-BIH database. According to the results, the suggested method gives predictions with an average accuracy of 97% in the first experiment, 99% in the second one, and 97.7% in the third experiment,without overfitting.

English Abstract

An electrocardiogram (ECG) can be dependably used as a measuring device to monitor cardiovascular function. The abnormal heartbeat appears in the ECG pattern and these abnormal signals are called arrhythmias. A faster and more accurate result can be reached by classifying and automatically detecting arrhythmia signals. Several machine learning approaches have been applied to enhance the accuracy of results and increase the speed and robustness of models. This research proposes a method based on Timeseries Classification using deep Convolutional -LSTM neural networks and Discrete Wavelet Transform to classify beats in three experiments, the first one is to classify 4 different types of Arrhythmia in the MIT-BIH Database. The second one for enhancement the first experimental results. The third one is for classifying the whole MIT-BIH database. According to the results, the suggested method gives predictions with an average accuracy of 97% in the first experiment, 99% in the second one, and 97.7% in the third experiment,without overfitting.

Item Type:	Thesis (Magister)
Identification Number:	042107
Subjects:	600 Technology (Applied sciences) > 621 Applied physics > 621.3 Electrical, magnetic, optical, communications, computer engineering; electronics, lighting > 621.38 Electronics, communications engineering > 621.381 Electronics
Divisions:	S2/S3 > Magister Teknik Elektro, Fakultas Teknik
Depositing User:	Unnamed user with username verry
Date Deposited:	25 Oct 2021 03:31
Last Modified:	24 Sep 2024 06:43
URI:	http://repository.ub.ac.id/id/eprint/185872

Text YOUSRA MOHAMMAD QASEM ALQAISI.pdf Download (4MB)

Actions (login required)

View Item