Comprehending the impact of deep learning algorithms on optimizing for recurring impediments associated with stress prediction using ECG data through statistical analysis

<p>Despite the myriad of stress related research studies, there has been very few studies which focused on the</p> <p>complexity of the ECG signal/data, prior to predicting stress. In order to counter the problem of “data complexity</p> <p>and overfitting”, we innovated ML (machine learning) approaches using transfer learning and autoencoder</p> <p>techniques, in order to predict stress/not stress (2 classes) with high precision from WESAD dataset. We then</p> <p>assessed the bias and variance associated with our algorithms through various statistical tests, in order to un-</p> <p>derstand their ability to generalize well on newer data. Our proposed algorithms were able to achieve 98.99%</p> <p>(CNN) and 98.92% (VGG16) accuracy through 10-fold cross validation, while maintaining a very low bias,</p> <p>variance, akaike information criterion (AIC) of and Bayesian information criterion (BIC) scores, substantiating</p> <p>their ability to predict stress with very high accuracy without overfitting. Results illustrates their ability to</p> <p>generalize well on any stress related data, irrespective of their complexity. Our algorithms performed better than</p> <p>every other related studies. Although we were able to reduce the time and space complexity of the algorithms</p> <p>through transfer learning and autoencoder techniques, the algorithms still require more time and computational</p> <p>power than simpler algorithms, something which will require more attention for our future work.</p>