Forecasting airline flight crew needs using machine learning and time series models

Although airline transportation is one of the sectors most affected by the Covid-19 pandemic, it is indispensable for the world economy. In recent times, economic crises have occurred from time to time due to epidemics and terrorist attacks. Although most airline companies in the world have downsized due to the effect of the pandemic, they continue to grow in the long run. With the introduction of the pandemic into our lives, cargo transportation in aviation has gained great importance as well as passenger transportation. In the face of increasing demands for all airline companies, the need for flight crew has been the main problem. Airline companies are carrying out various studies to solve this problem. The aim of this project is to determine the team need for the coming years by establishing a decision support system with the help of machine learning in order to correctly realize the following outputs. Along with the estimation of the number of crew needs, when variables such as production hours, number of aircraft and destination increase, flight operations will be able to be carried out full-time and completely in the coming years, Cargo, charter, etc. flights with additional income that will arrive at any time of the year will be carried out without any problems, by examining the usage rates of the number of sentry teams created based on the past, a more optimal number of shifts will be estimated for the future, thereby saving the cost of unnecessary shifts and ensuring a more optimal use of the existing teams. The project will consist of three successive basic steps. The first of these is to obtain data from the Crew Planning Presidency systems. The currently used planning system can present historical data to the planner in the form of reports. There is no need for any package program or library to obtain data. Another step is data analysis. Data mining operations will be performed using Python software language on the data obtained in this step. As a result of these processes, different machine learning algorithms will be applied on the data that will be ready for analysis. The algorithm that gives the most optimal result will be selected from the results obtained as a result of the applied algorithms. The last step of the project will be to create an interface for users. After the project is completed, the software will be made available to the planners. The relevant persons will estimate the number of teams required for the coming years through their studies and in line with this target, new teams will be employed, training will be planned for existing teams as needed, they will be promoted and the necessary needs will be met. The dataset and analysis outputs presented in this thesis have been modified, anonymized or represented symbolically in accordance with the Turkish Airline's confidentiality policies and ethical principles. The presented content has been prepared in a way that does not compromise the scientific validity of the study.