ECG Data Forecasting Based on Linear Models Approach: a Comparative Study

This paper investigates the replacement of the surface electrodes (physical sensors) measuring the electrocardiogram (ECG) signals by forecasting linear algorithms. The aim is to test the ability to overcome the loss of information in case of failure of any electrode. From real ECG signals measured in different auscultation sites, the ability to predict the ECG signal of one site depending on another site is evaluated by 3 methods. In this paper, based on quantitative criteria, a comparative study between Linear regression (LR) model, K-nearest neighbors model (KNN) and Random forest regression (RFR). The advantages and drawbacks of each one are also highlighted also that the three models are very accurate in building a new ECG signal.segment similar to the real signal.