Fundamental cognitive workload assessment: A machine learning comparative approach

Abstract

Mental workload remains an essential but challenging aspect of human factors, while machine learning serves as an emerging and expanding research realm to a wide variety of applications. This paper aims to comprehensively bridge the two areas by comparing present state-of-the-art machine learning approaches that are currently utilized for assessing cognitive workload, primarily artificial neural networks and support vector machines. To address and evaluate both approaches, we obtain a physiological data set used to study fear conditioning and cognitive load and format the data to focus primarily on the latter. Ultimately, the results indicate that both techniques can effectively model the data with up to 99% accuracy. Furthermore, under optimal parameter selection, the neural network model produces the highest possible accuracy under a comfortable level of deep learning while the support vector machine model employs greater speed and efficiency while still enjoying a respectably high level of accuracy.