The Airline Passenger Load Factor Prediction with Machine Learning - Linear Regression Method and Modeling with Fractional Calculus and Deep Assessment Methodology

Forecasting is used in air transportation to make short-term judgments and longer-term decisions to figure out how demand patterns will vary over time. In our paper, the passenger load factor (PLF) is forecasted using a variety of models so that a preliminary structure for both the economy and business classes can be obtained. To solve this problem, when each flight is assumed to have its own characteristics, a method that can reflect both the flight profile and the flight time dimension is needed. As a result, we used the least-squares method to create a continuous curve that is valid for any time interval by using the least-squares method with the 2015 daily reservation data of the 19-market areas of Turkish Airlines. Making predictions with machine learning was aimed at solving this large-scale problem in a computationally efficient way. Forecasting PLF is done by using information from two years of reservations, group sales data, calendar information, weekly dates, trend differences between the current year and the previous year, load factor information from the same period of the previous year and panel data to allow individual heterogeneity control. The findings of the fractional model when coefficient N is equal to 15 are approximately 2.4 times better than the DAM model and superior to regression analysis. With this study, it is expected to maximize revenue, modify capacity, increase flight route efficiency, optimize air traffic operation, and make projections for specific flying days in terms of revenue management