Hybrid Machine Learning–Statistical Method for Anomaly Detection in Flight Data

This paper investigates the use of an unsupervised hybrid statistical–local outlier factor algorithm to detect anomalies in time-series flight data. Flight data analysis is activity carried out by airlines primarily as a means improving safety and operation their fleet. Traditionally, this performed checking exceedances pre-set limits parameters. However, method highlights single events during flight, making laborious. The process also fails establish trends or reflect potential unknown hazards. research took advantage machine learning techniques recognize patterns large datasets implementing local (LOF). In order minimize human input, statistical approach was adopted threshold value above which flights are considered be anomalous interpret scores. shows that LOF quantifies degree outlier-ness rather than binary categorizing point into inlier outlier, case clustering algorithms. Thus, with LOF, for first time, we demonstrated aviation industry, could not only identified but given anomaly score compare two manner. Furthermore, helps track behavior time flight. insightful when abnormal, some seconds short duration. For attempted parameters responsible at least give direction experts looking cause abnormal behavior. all analyzed real-life manner contrast simulated