Using Machine Learning Techniques to Predict Block Time Padding and Reduce Flight Delays Ratio

This study leverages three machine learning models: Random Forest Regression, XGB Regression, and Gradient Boost Regression to optimize block time padding (BTP) for commercial flights using a public dataset from the Bureau of Aviation in the United States. Predictive features, such as flight duration, distance, taxi times, and aircraft-specific details, as well as unpredictable factors like variations in traffic during different times of the day and week, were used in the experiment. The best-performing machine learning model was Random Forest Regression, after achieving a Mean Absolute Error (MAE) of 3.017, a Root Mean Squared Error (RMSE) of 9.879, and an R-squared of 0.954 out of the three regression models.