Upsampling RR-Interval, is it Possible?

Vpsampling data on Electrocardiogram (ECG) is needed especially when obtaining data with a low sampling rate. This study aims to determine the effect of RR-interval (RRi) upsampling on the Heart Rate Variability (HRV) feature in the time domain, while testing the performance of several machine learning algorithms with the K-Fold Cross Validation approach. The study began by extracting RRi from ECG data through peak detection, then performing upsampling starting from 1.25x to 3x on RRi Origin. Each RRi data, both RRi Origin and RRi upsampling results, were then extracted with HRV values in the time domain. The results of HRV extraction were tested on several machine learning algorithms (Neural Network, Support Vector Machine, Decision Tree, K-Nearest Neighbor, and Gradient Boosting). The results show that overall RRi Origin (without upsampling) provides the best Accuracy, AUC, and F1-score on the Gradient Boosting prediction algorithm, at 78.54 %, 0.865, and 81.73%, respectively. However, 2.25x upsampling on Decision Tree achieves the best results in terms of Recall (91.81 %). This finding suggests that although RRi Origin tends to provide more balanced performance, upsampling at a certain level can increase the sensitivity of machine learning algorithms. This study shows that ensemble techniques can provide better accuracy for prediction. However, it is important to consider the characteristics of the dataset and the need for interpretability when choosing an algorithm.