Published Paper Details:

Accurately predicting flight delays is a key challenge in aviation analytics, with significant implications for airline operations and passenger satisfaction. This project presents a machine learning-based system for predicting flight delays using historical U.S. domestic flight data. The dataset is extensively preprocessed, including the imputation of missing values, encoding of categorical variables, and generation of time-based features such as scheduled departure hour and total delay duration. A binary classification approach is adopted, labeling flights as delayed when the arrival delay exceeds 15 minutes. Three models--Random Forest, Decision Tree, and Logistic Regression--are developed and evaluated. Both the Random Forest and Decision Tree classifiers achieved 100% accuracy on the test set, while Logistic Regression reached 98.4%, indicating the strong predictive capability of the selected features. However, such perfect accuracy suggests potential data leakage or overfitting, as the delay-related columns used in training directly influence the target variable. A user-friendly graphical interface, built with Tkinter, allows users to input flight number, tail number, origin, and destination to receive real-time delay predictions and reasons. Additionally, visual analytics generated using Seaborn highlight delay trends by departure hour and weekday. The system demonstrates how machine learning, combined with rich flight data, can support operational decisions in aviation, although care must be taken to ensure model generalization and reliability