Comparison of Support Vector Machine for Apron Allocation

Abstract

This paper presents machine learning techniques for classifying parking stand locations in the apron allocation management service that affects total airport ground service processing time at airports where arriving aircraft land. SVM and Kernel SVM algorithms will be used, as well as Polynomial, Gaussian RBF, and Sigmoid, based on five input factors: aircraft identification, estimated time of arrival (ETOA), area of apron, type of aircraft, and target of stands. Then, we compared classification accuracy and performance using the Mean Absolute Error (MAE) and the squared mean error (Root Mean Square Error: RMSE), and discovered that the Gaussian RBF kernel of the SVM algorithm model is more accurate than the other model. This work may be beneficial in assisting airport's decision-makers and enhancing airport operations efficiency and predictability.