ANALYSIS OF ARTIFICIAL INTELLIGENCE-BASED FLIGHT RISK ASSESSMENT AS A DECISION SUPPORT SYSTEM FOR ENHANCING NAVAL AVIATION SAFETY

Abstract

Naval aviation safety requires reliable assessment of interacting environmental, technical, human, and organizational risks that may not be adequately represented through conventional methods. This study aimed to analyze an artificial intelligence-based flight risk assessment model as a decision support system for enhancing naval aviation safety while preserving professional human judgment. A mixed-methods case study design was employed using 240 non-sensitive historical flight-safety records and questionnaire responses from 30 aviation personnel. The analysis combined machine-learning classification, descriptive statistics, inferential testing, semi-structured interviews, and triangulation. The results showed strong predictive performance, with 91.7% accuracy, 89.8% precision, 87.6% recall, 88.7% F1-score, and 0.93 ROC-AUC. AI-generated risk scores were strongly associated with established risk categories (r = 0.81, p < 0.001), while environmental conditions represented the strongest predictive factor. User assessments indicated high perceived usefulness and interpretability, although decision-support confidence was comparatively lower. The study concludes that artificial intelligence can strengthen naval aviation safety by integrating complex risk information and supporting evidence-based assessment. Effective implementation requires explainability, data quality, continuous validation, human oversight, and institutional governance to ensure that AI complements rather than replaces qualified aviation professionals. The findings support a human-centered safety architecture in which algorithmic outputs remain advisory, transparent, and contextually evaluated.