Towards objective human performance measurement for maritime safety: A new psychophysiological data-driven machine learning method

Human errors significantly contribute to transport accidents. Human performance measurement (HPM) is crucial to ensure human reliability and reduce human errors. However, how to address and reduce the subjective bias introduced by assessors in HPM and seafarer certification remains a key research challenge. This paper aims to develop a new psychophysiological data-driven machine learning method to realize the effective HPM in the maritime sector. It conducts experiments using a functional Near-Infrared Spectroscopy (fNIRS) technology and compares the performance of two groups in a maritime case (i.e. experienced and inexperienced seafarers in terms of different qualifications by certificates), via an Artificial Neural Network (ANN) model. The results have generated insightful implications and new contributions, including (1) the introduction of an objective criterion for assessors to monitor, assess, and support seafarer training and certification for maritime authorities; (2) the quantification of human response under specific missions, which serves as an index for a shipping company to evaluate seafarer reliability; (3) a supportive tool to evaluate human performance in complex emerging systems (e.g. Maritime Autonomous Surface Ship (MASS)) design for ship manufactures and shipbuilders.