Machine Learning-Based Prediction of Flight Punctuality and Delays

Air traffic demand has recently expanded considerably, creating worries about saturation, congestion, and queuing. Flight delays impact transportation networks, resulting in operational errors, higher economic losses for airlines, and consumer complaints and discontent. As a result, aviation authorities, universities, and researchers are increasingly interested in developing preventive techniques to mitigate the impact of aircraft delays. In this study, we created a predictive model to estimate airline delays for domestic flights in the United States through a survey of pilots, air traffic controllers, airport leaders, airport personnel, and passengers. Data for all the year 2018 were utilized with powerful Machine Learning algorithms such as feed-forward Neural Network, Random Forest, Gradient Boosting and Decision Trees. The conventional attributes that influence on flight delays were taken from the database used. To increase the performance of the suggested model, we created and introduced more variables that, as far as we know, had never been considered in prior research. As a consequence, with the incorporation of the conventional and suggested features at the same time, the model proved capable to estimate flight delays with an R-squared of 98.33% for the neural network.