Akademik Veri Yönetim Sistemi
JOURNAL OF THE FACULTY OF ENGINEERING AND ARCHITECTURE OF GAZI UNIVERSITY, cilt.39, sa.4, ss.2031-2040, 2024 (SCI-Expanded)
Management of unexpected events, which is the last stage of airline scheduling, has a very dynamic structure. This dynamic structure requires the effective use of available data in order to produce rapid solutions. For this purpose, machine learning techniques have been applied to analyze flight information and use it to help in decision making. In this paper, machine learning techniques were used to classify flights according to delay categories and find critical facors for minimizing delays. These classification methods are suppurt vector machine, random forest, C 5.0, k -nearest neighbor and CART algorithms. An example for the comparing prediction success of these methods in finding the flights with the most delays (abnormal filights) are shown in Figure A. In the Figure A, the x-axis shows the amount of samples considered and the y-axis shows how many of them were predicted correctly. Additionally, the best possible prediction curve, apart from the algorithms used, is also shown in blue at the top. The red line in the middle indicates how much would be correct even if a random choice was made. As can be seen from Figure A, the 'Abnormal Delay' class was predicted most correctly in support vector machine (yellow curve) and random forest (green curve). Purpose: The purpose of this study is to classify flights according to their delays and find important delay factors in order to help the decision maker to produce quick solutions for disrupted flights in airline disruption management. Theory and Methods: In the study, in there scheduling process related to flight disruptions; machine learning algorithms were used to determine the risk factors, to easily access meaningful data and to make a delay prediction for flights to help decision making. For minimizing flight delays and helping decision maker to produce quick solutions, classification is considered and flights are classified according to their delay time. Results: The aim of the study is to provide a preliminary idea about the flights to the decision-maker by ensuring that the flights are evaluated according to the delay classes. Accordingly, it has been seen that predicting the delay classes of flights will help the decision-maker in terms of providing new information on disruption management, which has been an important research topic in recent years. Then the use of different machine learning techniques for delay prediction has been studied. Conclusions: Also, delay prediction is carried out in the literature in two different ways as robust scheduling and schedule recovery. It has been observed that it has a structure that carries the characteristics of both. For this reason, it is thought that accurate delay prediction will be beneficial to operators. [GRAPHICS] With the increasing competition in the aviation industry in recent years, airline companies have tended to manage their operations more efficiently. Airline scheduling activities are carried out in four stages as flight scheduling, air craft scheduling, crew scheduling and disruption management. In the first three stages, a feasible flight schedule is created for the system, and in the last stage, solutions are sought for the problems that arise during the flight. Airline companies face a serious loss of time and cost constraints in flight disruptions. The most difficult aspect of flight disruption management is the necessity of rescheduling the plans in minutes. When flight disruptions occur, in the case of rescheduling, companies make decisions based on traditional methods, in tuition or experience, and the excess of details in operations negatively affects the decision. In the study, in there scheduling process related to flight disruptions; machine learning algorithms were used to determine the risk factors, to easily access meaningful data and to make a delay prediction for flights to help decision making. Machine learning has provided support to the decision maker in rescheduling by extracting meaningful new information from the current data, and flight delays have been predicted with classification algorithms.