Machine Leaning Based Approach towards Secured Aviation Safety Management Systems: A Case study of Tanzania Civil Aviation Authority

Abstract

Aviation safety remains a global priority due to its direct impact on lives, economic stability, and public confidence in air transportation. Even though the aviation safety management systems play a very crucial role, most of the existing manual-based safety management systems face challenges, including under-reporting, manual processing inefficiencies, and limited predictive capabilities.  In an attempt to address the challenges, this study introduces a machine learning (ML) based approach that enhances aviation safety management systems, including elements of safety hazard reporting and safety risk assessment. The proposed approach automates the prediction of safety hazard likelihood, severity, and risk levels assessment. It provides intelligent recommendations for proactive safety management before an incident or accident occurs. The study has used the five-year datasets (2019-2024) from the Tanzania Civil Aviation Authority (TCAA). The proposed ML approach utilises Linear Support Vector Classifier (Lear SVC), Gradient Boosting and Random Forest, which were integrated with the developed Electronic Safety Hazard Reporting Portal (eSHRP). Evaluation of the proposed approach indicates that the ML approach has improved accuracy by 90% in safety hazard assessment, enhanced safety decision-making, and expedited safety hazard response through automated Short Message Service (SMS), Email notifications, and Visual analytics.