PREDICTIVE FLIGHT DELAY ANALYSIS SYSTEM USING ML TO PROVIDE TRAVELERS WITH REAL TIME UPDATES AND RESCHEDULING OPTIONS

Abstract

Keeping flight disruption to an acceptable level is not a matter of just reporting when it happens. This implementation paper presents a machine-learning based flight delay analysis platform that evaluates delay risk based on the operational flight parameters and converts the prediction to a practical travel assistant. The implementation pipeline comprises of validation of data, preprocessing of data, categorical encoding, feature preparation, model training, comparative assessment, prediction, notification, and rescheduling support. Instead of having a single learner, LR, Decision Tree Regression, Bayesian Ridge, Random Forest Regression and Gradient Boosting Regression are studied to compare the behavior of the models. Application separates administrative model management tasks and passenger facing prediction tasks and exposes selected model through a web interface. The predictive inputs may include flight number, carrier, route, schedule, temporal features, distance, weather indicators, congestion and/or previous delay trends. The result of the implementation will be a delay status or estimate delay, the result in human readable format, and alternate scheduling advice when there is a significant chance of interruption. The study offers a sound architecture for implementing predictive analytics and a decision support system for passengers, and maintains flexibility for integration with real-time aviation and meteorological information in the future. “Key Words: Flight Delay, ML, Predictive Analytics, Random Forest, Gradient Boosting, Real-Time Updates, Rescheduling, Aviation Decision Support”