Advanced K-means clustering algorithm for large ECG data sets based on a collaboration of compressed sensing theory and K-SVD approach

Clustering of long-term recording electrocardiography (ECG) signals in the healthcare systems is the most common source in detecting cardiovascular diseases as well as treating heart disorders. Currently used clustering algorithms do have their share of drawbacks: (1) Clustering and classification cannot be done in real time; (2) Implementing existing algorithms would lead to higher computational costs. These drawbacks motivated us in developing novel optimized clustering algorithm which could easily scan large ECG datasets for characteristic biomarkers. In this paper, we present an advanced K-means clustering algorithm based on compressed sensing theory in combination with the K-singular value decomposition method. We validate the proposed algorithm's performance with principal component analysis and linear correlation coefficient dimensionality reduction methods followed by sorting the data using the K-nearest neighbors and probabilistic neural network classifiers. The proposed algorithm outperforms existing algorithms by achieving a classification accuracy of 99.98% (increasing 11% classification accuracy compared to the existing algorithm). This ability allows reducing 15% of average classification error, 10% of training error, and 20% of root-mean-square error. The proposed algorithm also reduces 13% clustering energy consumption compared to the existing clustering algorithm by increasing the classification performance.