APPLICATION OF MACHINE LEARNING METHODS FOR PREDICTION OF SEAFARER SAFETY PERCEPTION

Abstract

Purpose - This study aims to predict seafarer safety perceptions and evaluate their feedbacks in order to understand the human factor on ship’s safety. Design/methodology/approach - A questionnaire survey has been conducted with 304 seafarers' participation and they responded several safety climate and perception indicators that based on literature, for instance safety assessment of supervisors and company, company's training arrangement, accident and near miss reporting etc. Scores of survey results have been estimated with four machine learning algorithms, namely as multiple linear regression, support vector regression, random forest and decision tree regression. Findings - The multiple linear regression method gave the best prediction performance for seafarer safety perception level with 4.07 mean absolute percentage error. Originality - It was seen that the machine learning techniques can be applicable in the prediction of seafarer safety perception based on collected data. This study may provide useful perspectives for maritime companies in the improving safety on ships.