Decision-making in fuel tankering with machine learning techniques in aviation and its application

In this thesis, it is aimed at decision-making in fuel tankering with machine learning techniques in aviation. The dataset used in this study is obtained from a commercial airline company based in Turkey. There are several basic customizable formulas/models used in the fuel tankering calculation referred in the literature. In this study, it is aimed to predict fuel tankering in the airline industry with machine learning algorithms that learn from raw data which is independent of these formulas/models. In this study, alternative machine learning models are created to classify the fuel tankering status as "Yes/No" and to predict the amount of fuel tankering. k-Nearest Neighbour algorithm, Naive Bayes Classifier, Classification and Regression Trees, Random Forest algorithm, Artificial Neural Networks, and Support Vector Machines algorithm are used. Performance of classification and regression models are compared with each other. Analyzes are carried out in RStudio with the R programming language. An online Shiny application is developed with the best performing Random Forest classification (Accuracy=0,9646) and regression (R²=0,8104) models.