Performance Evaluation of Machine Learning Algorithms in Estimating Taxi Times at Istanbul Airport

Abstract

This article evaluates the performance of regression algorithms used to estimate taxi-out times at Istanbul Airport. Artificial neural networks, random forest, gradient boosting, and decision trees algorithms were studied to determine the algorithms with the highest accuracy. Principal Component Analysis (PCA) was used to reduce the data's dimensionality and improve model performance. The findings of the study provide valuable insights for more effective management of airport operations and reduction of flight delays. PCA-applied Artificial Neural Networks (ANN) emerged as the most successful algorithm, demonstrating the highest accuracy (R²: 95.89%) and lowest error margins (MAE: 0.016, MSE: 0.001) in predicting taxi-out times. This superior performance indicates that ANN can effectively capture the complex relationships and variability inherent in airport operational data. Following ANN, the PCA-applied Random Forest algorithm also showed commendable accuracy (R²: 94.89%), providing robust predictions with slightly higher error margins (MAE: 0.157, MSE: 0.044) compared to ANN. These results underline the potential of using advanced machine learning techniques to enhance the efficiency of airport operations, thereby minimizing delays and optimizing resource allocation. Overall, the application of these machine learning models, particularly ANN and Random Forest, offers a significant improvement over traditional methods. The study's outcomes suggest that incorporating these advanced algorithms can lead to more accurate predictions of taxi-out times, supporting better decision-making processes and operational strategies at airports.