Iterative machine and deep learning approach for aviation delay prediction

In the aviation industry, flight arrival delays cause approximately 18 billion of loss to customers as stated in the literature. So, it becomes inevitable on the part of the aviation authorities to predict such delays and take necessary action to fix this loss for customer satisfaction. In this paper, an approach based on machine learning techniques is proposed that predicts the flight arrival delays considering input parameters ranging from distance to their corresponding weather details to make a decision of whether the specific flight is delayed or not. It makes use of neural networks and deep learning concepts to estimate flight delay. The proposed approach is tested on real world flight big dataset that gives an accuracy of 77% using deep nets and 89% using neural nets. This approach can achieve reliable prediction with respect to if flight arrival delay is to be expected or not, moving forward the use of such a model can come in handy not only for airline administrators but also the passengers who can rearrange their schedules and arrange accommodation.