Machine Learning Techniques to Analyze Operator’s Behavior

Abstract

With more effective management teams, airlines are becoming more stable, more productive, and more punctual. The problems plaguing the aviation industry, however, have not gone away, and instead they have become more complicated. Schedule recovery is the process of recovery operating disturbances. The operator can either solve the problem manually, use a solution created by the recovery solver, or use a combination of both. The recovery solver from Jeppesen is a software tool that produces a set of solutions to resolve these operational disruptions. This research has been carried out at Jeppesen, a Boeing company. To analyze the Jeppesen airline system and recovery solver extensively and to identify various machine learning algorithms that can be used to answer the following questions: "Will, the operator, use the recovery solver?" and "If the operator uses the recovery solver, which solution will she prefer?" In this paper, we thoroughly study and understand the historical labeled data of alerts from a Mexico-based airline company created during disruptions. We have labeled the data points into two categories: manual solution and recovery solver solution. The experimental results obtained from this project have shown that that neural network models do not significantly improve predictive performance compared to the boosting models.

Keywords

• airline disruptions
• machine learning
• supervised learning
• neural networks
• schedule recovery