Abstract
Healthcare systems recognize and are trying to alleviate a common and rapidly growing psychological illness called stress. Stress detection is becoming the most valuable task of the healthcare industry. Thus, much research is being conducted in this domain. With the wide use of chest- and wrist-worn devices as part of the smart healthcare system utilizing emerging information technologies such as big data, the Internet of Things, and artificial intelligence, it is becoming easier to collect the relevant data and to interpret them correctly. This research aimed to evaluate and analyze the popular stress detection dataset Wearable Stress and Affect Detection (WESAD) using the RandOm Convolutional KErnel Transform (ROCKET) technique. Such technique was selected due to its ability to extract multiple features for the classification process without losing important data. The classification process utilizes the linear ridge classifier, and the results obtained from the use of the ROCKET technique in this study were compared to the previously published results of the successfully performed previous studies in this domain. The ROCKET technique showed satisfactory performance with 87% accuracy, close to and within the range of the results obtained from the algorithms that were previously applied on the WESAD dataset. The study results show the great potential of the ROCKET technique, which can be further improved by utilizing different existing classifiers or by proposing new models based on ridge or logistic regression.