An AI-based Model for Predicting Flight Delays to Enhance Air Traffic Operations

Abstract

Flight delays pose ongoing challenges in aviation, impacting operations and customer satisfaction. This thesis introduces an AI-based forecasting model using historical flight data from Kaggle, including weather, delay causes, and operational metrics. After preprocessing—outlier removal, imputation, and feature engineering—feature selection was performed using SelectKBest and RFE. A Random Forest Classifier, optimized via GridSearchCV, achieved 98.41% accuracy with RFE-selected features. Key predictors included weather and carrier-specific delays. Visual analyses supported the findings. The study demonstrates AI’s effectiveness in predicting delays and sets the stage for future real-time applications using cloud infrastructure and live data to enhance flight operations.