Quantifying Bird Strikes due to Migration from a Machine Learning Perspective

Bird Strikes have always been a persistent threat to the aviation industry. Bird strikes occur when a bird at flight accidentally collides with an aircraft. Such collisions, be it a major or minor ones; often cause catastrophic structural damages and financial losses to both the aircraft and the aviation sector. From the earliest days of aircrafts, birds have always been a source of risk due to the common means of locomotion. Hence, this calls for nuanced approaches and strategies for mitigation of such incidents. Although there exist solutions that have real-time applications, it is important to note that none of these solutions are eco-friendly and environmentally sustainable. Adding to that, there is no solution that considers the correlation factor between migration and such incidents. Additionally, it was observed that out of 442 unique species in bird strike incidents through the dataset acquired for this study, 369 species were migratory leading to a conclusion that majority of the species involved in such incidents are migratory. Therefore, going to the depths of this issue to help prevent bird strikes while proactively working is more important rather than focusing solely on reactive mitigation strategies that are mostly taken post the detection of bird activity. Thus, through this focused study, a thorough understanding of correlation between bird strikes and migration has been established by utilizing the eBird migration data sourced from Cornell Lab of Ornithology and the bird strike data from Federal Aviation Administration (FAA). By leveraging an ensemble of models like Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Recurrent Neural Networks (RNN), Long-Short Term Memory (LSTM) and Support Vector Machines (SVM), migratory months for all the species were predicted. A correlation of about 53% was achieved between bird strike incidents and migration cycles. This conclusion can be used to ensure better safety measures at the airports prior to flight scheduling and many more other applications.