Key result
A Perceptron neural network utilizing feature reduction techniques achieved 100% classification accuracy for cardiac arrhythmia using only 14 features, compared to 92.324% with all 279 features.
Why the study?
Heart diseases cause the maximum death worldwide, prompting the use of cardiac arrhythmia databases and neural networks to improve the speed of processing cardiac signals with minimum errors.
Applying feature selection and genetic algorithms to ECG and clinical data can optimize neural network performance for arrhythmia classification, achieving high accuracy with fewer variables.
Answers are grounded in this paper and cite their sources.
Optimized feature selection may streamline neural network arrhythmia classification; leaves open prospective clinical validation.
Nowadays, heart diseases cause the maximum death in the world. Also, due to the noticeable increase of heart diseases, studying this field is one of the important matters in medical community. Therefore, this study tries to benefit using information in data base of cardiac arrhythmia and employ arterial intelligent and neural network, in order to improve the speed in getting cardiac signals with minimum errors and maximum certainty. The dataset for the project is taken from the UCI machine learning repository https://archive.ics.uci.edu/m1/datasets/Arrhythmia. Used data base has 279 characteristics taken from 364 patients that includes general characteristics and ECG signals received from patients. In this study, firstly the primary classification has done with all characteristics, so some parts of information in data base of cardiac arrhythmia with values near zero has omitted. Considering the improvement of accuracy of the classification after omitting the characteristics near to zero, in next level the second series of data in neural networks that has negative effect on classification has omitted. In order to increase the accuracy of neural network and minimize the number of characteristics, the characteristics has classified in multiple classes and the obtained ratio has improved using genetic algorithm. In this level, the best accuracy of the neural classification has obtained but in order to get a network with minimum characteristics possible and preserve the 100% accuracy of the classification, ineffective characteristics has omitted using PCA algorithm.
No takes yet. Share an insight, caveat, or question.
Jafari et al. (2020) studied Cardiac arrhythmia (n=364). Perceptron neural network with feature reduction (FDR, genetic algorithm, and PCA) vs. Perceptron neural network using all 279 features was evaluated on Classification accuracy. A Perceptron neural network utilizing feature reduction techniques achieved 100% classification accuracy for cardiac arrhythmia using only 14 features, compared to 92.324% with all 279 features.
Synapse has enriched 10 related papers. Consider them for comparative context: