Samenvatting

Aviation safety relies greatly on understanding human factors and pilot behaviour. This study uses innovative eye-tracking biometrics and predictive analytics to enhance aviation landing performance in a twin-engine Piper Seneca simulation. With a Tobii Glass 3 eye tracker, gaze patterns from two pilots with successful and failed landings were recorded, preprocessed, and engineered into temporal features in 5-s sliding windows. Analysis showed a clear visual strategy and specified that Successful landings had much longer fixation times, larger saccade amplitudes, and greater vertical gaze change. This indicates a focused approach to important cues. In contrast, failed landings had shorter fixations and smaller saccades, which showed a less coordinated visual strategy. Logistic Regression models trained on raw, mapped, and merged gaze data achieved high prediction accuracies between 0.91 and 0.97, along with ROC AUC values from 0.94 to 0.95. This allowed them to differentiate between successful and failed landings easily. Mapped gaze data were the strongest predictors, especially mean vertical gaze position and saccade amplitude. This highlights how important spatial context is for performance evaluation. A prediction simulation under real-time conditions demonstrated the model’s ability to monitor in real-time during flight training. Although the dataset was limited, these promising results suggest eye-tracking biometrics could help identify the best visual strategies and provide real-time feedback. This study can change the pilot training and improve aviation safety. Moreover, future research will focus on a large sample of pilots and combine with advanced temporal models. The ultimate goal is to maximise generalizability and practical application.