Anomaly Detection in Flight Data Using the Naïve Bayes Classifier

Abstract

Safety is the key to reliable civil aviation. In the airline industry, there is a growing emphasis on proactive safety management systems in order to improve the safety of current aviation operations. These systems utilize anomaly detection techniques to recognize and reduce the risk of accidents occurring. This work develops a new anomaly detection approach for commercial flight operations using routine operational data to enhance proactive safety management systems and utilizes data mining techniques to identify abnormal situations instantaneously during flights using real-life FDR (Flight Data Recorder) data. The Naïve Bayes classifier was used to detect normal and abnormal situations. This classifier was applied to a dataset of 100 flights and new abnormal situations could be recognized with a high probability of detection and a low probability of false alarm. The results strongly suggest that anomalies detected in a variety of flights can be recognized, which can help airlines with many different approaches, such as the deployment of predictive maintenance, the detection of early signs of performance divergence, safety support, and the training of staff accordingly.