Application of machine learning techniques for incident-accident classification problem in aviation safety management

Abstract

In a global world, one of key contributors to business success is the aviation industry. Most business magnets rely on this to quickly travel from one place to another to conduct transactions. This has led to an increase in demand of flights with many more aviation entrepreneurs cropping up every now and then. A key challenge faced by this industry is transporting passengers and goods safely from departure point to their destination. It is not uncommon for flights to encounter an incident/accident problem which often perturbs the journey. Thus, companies continuously strive to mitigate such occurrence. Aviation companies record historical data in various databases which is rich with insight that can be extracted and used to mitigate events that impede flights. In this research, we employed machine learning techniques to identify potential contributory factors and trained machine learning models using supervised learning. These models learn from existing data and are used to predict when an aircraft incident or accident would occur given some conditions. In our findings, the best model achieved an accuracy of 96% when tested on unseen data.