A Novel Evaluation of Flight Delay Analysis for Lithuanian Airports Using Supervised Machine Learning Based Boosting Techniques

Due to its reputation as the quickest mode of transportation, air travel has earned its passengers’ trust over the years. However, the airline industry has had to adapt to the reality that flight arrival delays are inevitable due to the fact that they are directly related to the way airspace and runways are managed. For the aviation industry to become more efficient, accurate prediction of flight delays is essential. Recent research has focused on developing methods for employing supervised machine learning to anticipate when flights may be delayed. The paper examines the variation in flight times between airports in Lithuania. The SMOTE method is used to achieve data parity. The latest FCMIM method was applied for feature selection. To forecast the time delay deviation of future bouts, a supervised machine learning model has been constructed. Tree boosting techniques (XGBoost, LightGBM, and AdaBoost) have been used for the investigation. Each algorithm’s performance has been quantified across four dimensions: recall, precision, F1-measure, and accuracy. The freshly gathered dataset from Lithuanian airports and meteorological data on departure/landing time has been used for every experimental inquiry. Fights at both the airport and the port have been studied independently. The results show that the boosted trees approach is the best predictor of tree model classifiers, which have the highest accuracy. Accuracy on the Departure Dataset is 98% for the suggested models, while accuracy on the Arrival Dataset is 91%. When compared to other approaches, its accuracy is clearly superior. The proposed models are able to minimise over-appropriation and increase forecast accuracy.