Automated Arrhythmia Classification using Harris Hawks Optimization with Deep Learning Model on IoT Environment
S. Jagadeesh, Manish Shrimali, Pvs Siva Prasad, K. Maheswari, A. Narendrakumar, G. Koteswara Reddy
Abstract
An electrocardiogram (ECG) has been widely utilized for evaluating heart disease which is effect more explores cardiac problems and recognition of cardiovascular diseases. The procedure is easy, fast, and non-invasive. But, artificially identifying and classifying heart disease is difficult, as manually examining the ECG signals is time-consuming. In recent years, a huge growth in the control of IoT can witness that gives a lot in the health care method as it allows continuous monitoring of patients but there is a requirement for advanced automatic monitoring methods for classifier of cardiac arrests. This article develops an automated arrhythmia classification using Harris Hawks Optimization-based DL (AC-HHODL) technique in an IoT environment. The AC-HHODL technique focuses on the identification and categorization of arrhythmia using ECG signals in the IoT environment. In the AC-HHODL technique, the MobileNetv2 model was executed for producing a collection of feature vectors. In addition, the HHO technique was exploited for the optimum hyperparameter adjustment of the MobileNetv2 model. Furthermore, least square support vector machine (LS -SVM) approach is applied for accurate arrhythmia classification. To validate the better performance of the AC-HHODL approach, the experiments were carried out on arrhythmia dataset. On the other hand, with respect to sens <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">y</inf> , the AC-HHODL algorithm has managed to report increased sens <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">y</inf> of 98.65% while the ResNet, ResNet-attention, DenseNet, VGG-16, and XGBoost approaches have exposed degraded sens <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">y</inf> of 91.67%, 89.47%, 95.34%, 95.77% and 90.40% respectively. The simulation output inferred the improvised efficacy of the AC-HHODL technique over other models.