Publicação
Flight arrival delay prediction in European Airports: A Machine Learning approach structured by CRISP-DM
| Resumo: | The continuous increase in air transport demand across Europe has contributed to growing delays at airports, negatively impacting operational efficiency and passenger experience. Pre-dicting arrival delays emerges as a strategic tool to mitigate these effects. This study proposes a predictive model based on machine learning techniques, structured according to the CRISP-DM methodology, with the goal of anticipating both the occurrence and the duration of arrival delays at major European airports from 2015 to 2019. The approach is implemented in two stages: the first applies classification models to identify flights with a high probability of delay (over 15 minutes); the second estimates the delay duration using regression models. The Mul-tilayer Perceptron (MLP) algorithm showed the best performance, with 88% accuracy, 87% recall, and an AUC of 0.87 in the classification task. In the regression stage, it achieved an R² of 73% and RMSE of 10.42 minutes, while XGBoost recorded the lowest mean absolute error (MAE) of 6.11 minutes. Feature importance analysis highlighted total flight time and departure delay as the most significant predictors, along with other relevant factors such as distance, month, and quarter. The results are consistent with previous studies and reinforce the potential of machine learning techniques in improving air traffic management. This research contributes to the field of Business Analytics by proposing an interpretable and effective model to support operational decision-making in the European aviation context. |
|---|---|
| Autores principais: | Farias, Myslane Kalyne de |
| Assunto: | Flight arrival delay Machine learning -- Machine learning CRISP-DM European airports Business analytics Atraso na chegada de voos Aprendizado de máquina Aeroportos europeus Análise de negócios |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso restrito |
| Instituição associada: | ISCTE |
| Idioma: | inglês |
| Origem: | Repositório ISCTE |
| Resumo: | The continuous increase in air transport demand across Europe has contributed to growing delays at airports, negatively impacting operational efficiency and passenger experience. Pre-dicting arrival delays emerges as a strategic tool to mitigate these effects. This study proposes a predictive model based on machine learning techniques, structured according to the CRISP-DM methodology, with the goal of anticipating both the occurrence and the duration of arrival delays at major European airports from 2015 to 2019. The approach is implemented in two stages: the first applies classification models to identify flights with a high probability of delay (over 15 minutes); the second estimates the delay duration using regression models. The Mul-tilayer Perceptron (MLP) algorithm showed the best performance, with 88% accuracy, 87% recall, and an AUC of 0.87 in the classification task. In the regression stage, it achieved an R² of 73% and RMSE of 10.42 minutes, while XGBoost recorded the lowest mean absolute error (MAE) of 6.11 minutes. Feature importance analysis highlighted total flight time and departure delay as the most significant predictors, along with other relevant factors such as distance, month, and quarter. The results are consistent with previous studies and reinforce the potential of machine learning techniques in improving air traffic management. This research contributes to the field of Business Analytics by proposing an interpretable and effective model to support operational decision-making in the European aviation context. |
|---|