EKG verilerinin destek vektör regresyon yöntemiyle sıkıştırılması

Abstract

Highlights:Graphical/Tabular Abstract Purpose: Electrocardiogram (ECG) signals must be continuously recorded and monitored to effectively detect diseases caused by fast or slow heartbeat, that is, rhythm disorders. However, long monitoring periods generates large amount of data that are difficult to store and transmit. Moreover, these records may be subject to noise due to the environment. For this reason, an effective and reliable data compression technique is needed for ECG data compression without losing the clinical information content. In order to address the aforementioned problems, in this paper, SVR based compression algorithm is introduced. Theory and Methods: SVR provides a better generalization capability because it tries to minimize both the empirical risk minimization and the structural risk minimization principle. Therefore, SVR has been widely used in many scientific areas. The success of SVR in application is dependent on the ε-insensitive Laplace (Vapnik) loss function that ignores the errors lower than the user-defined ε value. Small noisy training samples falling into the ε-insensitive zone are not included in the solution representation. So, SVR yields a sparse (compressed) model in the obtained solution. The compressed signal is expressed as the weighted sum of basis functions. Different from the other transformation-based compression methods, the number, position, and shape of these functions are automatically determined by the SVR algorithm based on the solution of the quadratic optimization problem. Results:From the experimental results, the user defined ε parameter allows us to check the selection of the samples (support vectors) related to the direct compression. If the value of the ε parameter increases, the Compression Ratio (CR) also increases, and at the same time increases the Percent Root Mean Square Difference(PRD) and Root Mean Square Error (RMSE) values, which leads to distortions in the compressed signal. Also, when the value of ε increases, the SVR algorithm also increases in generalization ability (lowers w). Conclusion:Computer simulation results show that the performance of the proposed SVR-based compression algorithm is better than other transformation-based compression techniques such as the commonly used Fourier Transform (FT), Discrete Cosine Transform (DCT), and Discrete Wavelet Transform (DWT). As can be seen from the table, at the same compression ratio (SO = 6.38), the SVR technique reached minimum PRD = 17.30 and RMSE = 0.021. In addition, the proposed method has important features such as being independent of the sampling conditions, being a single minimum in the global sense, and not requiring additional algorithms (preprocessing of the ECG signal). As a result, SVR based compression algorithm is an attractive candidate for compressing ECG signals.

Hybrid Modeling for Stream Flow Estimation: Integrating Machine Learning and Federated Learning

“…Within the ambit of our study, we further augmented the SVR model's capabilities by blending it with the formidable radial basis function (RBF) kernel. This blending process is instrumental in calculating the proximity and similarity between two data points, a critical aspect that underpins SVR's predictive capabilities in our research endeavors [58][59][60]. The judicious combination of SVR and the RBF kernel fortifies the model's capacity to capture intricate patterns and relationships within the data, culminating in a potent tool for accurate prediction and forecasting in our study.…”

ECG Data Compression Using ε-insensitive Quadratic Loss Function

“…Recently, unlike current transform based compression techniques, Karal [10] has proposed a new method based on standard support vector regression (SVR) to compress ECG data according to the given error tolerance in an optimal and rapid manner. It has also been shown that the performance of the proposed standard SVR-based compression technique is higher than that of transform-based compression techniques such as FT, DCT, and DWT commonly used in the literature.…”