A classifier of aircraft trajectories based on several subsets of predictors, such as temporal, geometric, and meteorological data, is proposed in this paper. This research utilizes a use case scenario of $294$ trajectories departing from Zagreb Pleso airport with London Heathrow as the destination airport. Manual labeling was used to determine two classes based on position in the third point of a trajectory. Classification methods included the k-Nearest Neighbours (k-NN), Gaussian Process (GP), Decision Tree (DT), Random Forest (RF), Multilayer Perceptron (MLP), Naive Bayes (NB), Quadratic Discriminant Analysis (QDA), AdaBoost (AB), and Linear and Radial Basis Function (RBF) Support Vector Machine (SVM) algorithms. The GP method produced a $97.85\%$ testing accuracy by employing the arithmetic average of diffusion distance and direction change as predictors. Using all evaluated trajectory features, the AdaBoost approach has the same performance with an almost ten times longer execution time, which makes it less appropriate for small, low-performance, and inexpensive portable systems like smartphones. The theoretical definitions of direction change and diffusion distance, which use vector and scalar values to represent both dynamic and static features of the trajectory, support the decision to include them in the final model.
