Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/0C6B2AE5-CFB0-41B2-8B51-C772A2DB63CD | ||
| Year | 2024 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Alexandros Chatzipetros, "Predicting the occurrence of flight delay based on machine learning techniques", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2024 https://doi.org/10.26233/heallink.tuc.101279 | ||
| Appears in Collections |
In recent years, there has been a great deal of interest in flight delay prediction, since flight delays have a negative impact on both the economy and the environment, as they increase fuel consumption and therefore carbon emissions. Statistical and operational methods for prediction have been employed in this area, and with the advent of technology, Machine Learning techniques serve also a catalytic role in the study and forecasting of flight delays. Prompted by the aforementioned facts, in this Diploma Thesis we applied Machine Learning methods to forecast a) the average departure delay of a flight, b) the average delay throughout a flight, and c) the average total delay of a flight, using real-world, chronological data. In addition, two strategies were used to estimate the overall delay: either by aggregating the findings of the departure delay and flight delay prediction models, or by applying a single model to predict the total delay directly. Traditional Machine Learning approaches, such as Linear Regression, Polynomial Regression, and Support Vector Regression, as well as Neural Networks, were employed to develop our predictions. To determine the efficacy of the algorithms, the Mean Absolute Error and Root Mean Square Error metrics were employed, along with the R-Squared Determination Coefficient metric, in conjunction with the Cross Validation approach, while the execution time of each implementation was also considered. In addition, the effect of a new heuristic factor, namely the knowledge of previous delays at both the arrival and departure airports, was examined. In conclusion, the application of our methodology to annual flights between two of the busiest international airports in the United States of America (USA) demonstrates that the use of Machine Learning algorithms can contribute to flight delay prediction, while the novel feature of knowing previous delays positively affects the accuracy of the predictions.
Copyright © DIAS 2013
