Özet:

The Covid-19 epidemic, which emerged in China in 2019, spread all over the world in a short time. Due to the epidemic, many sectors around the world have been adversely affected. Airline passenger transport is also one of the sectors most heavily affected by the Covid-19 outbreak. Airline passenger mile is a frequently used metric in the aviation industry and is obtained by multiplying the number of miles flown by the number of passengers. With the airline passenger mile metric, mobility in the sector can be measured. In this study, a total of 259 data for the American airline passenger mile metric taken from the United Reserve Economic Data (FRED) system between 2000 and 2021 were used. Since the data set used has an upward trend, it does not show a static feature. In addition, it has been observed that it has a seasonality that increases in some seasons of the year and decreases in some seasons, on the contrary. Long Short-Term Memory (LSTM) architecture of Recurrent Neural Networks (RNN), which is one of the deep learning methods, was preferred in this study because it can work with non-stationary data set and can take seasonal effects into account. In order to observe the effect of the sudden change in the trend of the data set during the Covid-19 period on the performance of the LSTM architecture, two separate estimations were made with both the data set containing the Covid-19 period data and the data set not containing the Covid-19 period, and the results were compared. According to the findings, it was seen that the performance of the LSTM architecture was much higher in the predictions made with the data set that did not include the Covid-19 period. The same data set was also estimated with the Autoregressive Model and compared with the performance of the LSTM architecture. Finally, with the LSTM architecture, which gives more successful results, estimations were made with the number of passengers in Turkey between the years 1960-2020.

Anahtar Kelimeler: