Application of Transfer Learning in Continuous Time Series for Anomaly Detection in Commercial Aircraft Flight Data

Abstract

In recent years, transfer learning has attracted widespread attention and research. Transfer learning is a new machine learning method that uses existing knowledge to solve different but related domain problems. It relaxes two basic assumptions in traditional machine learning: (1) Training samples for learning and new test samples satisfy the conditions of independent and identical distribution; (2) There must be enough training samples available to learn a good model. Since it is costly and dangerous to repeat testing flights at extreme conditions, building an anomaly detection model for aircraft flight is also constrained by insufficient samples in limited data for different testing flight scenarios. To handle the insufficient samples, we propose a transfer-learning based approach to establishing an anomaly detection model for dangerous actions of aircraft testing flights. In our approach, we transfer the "knowledge" obtained in some testing scenarios to other scenarios containing dangerous action. Evaluation results indicate that our approach works well in terms of convincing accuracy in prediction by models in target scenarios transferred from source scenarios.