FORECASTING THE MONTHLY AIR PASSENGERS WITH RNN, GRU AND LSTM METHODS IN THE COVID-19 PERIOD


Creative Commons License

Karaboğa H. A. , Demir İ., Çelik R.

Engineering (ICMASE 2021) , Salamanca, Spain, 1 - 12 June 2021, pp.87-88

  • Publication Type: Conference Paper / Summary Text
  • City: Salamanca
  • Country: Spain
  • Page Numbers: pp.87-88

Abstract

The covid-19 pandemic has negatively affected the transportation and tourism sectors. Although the magnitude of this effect can be reduced by various measures, it is estimated that it will last for many years. Since the countries blocked the entrances and exits to prevent the spread of the virus, there have been significant decreases in the number of air passengers. As countries blocked the entrance and exit to prevent the spread of the virus, there have been very significant decreases in the number of air passengers. Turkey, which is one of the most important transit points, was particularly affected by this situation. It is necessary to forecast the number of air passengers in order to manage the crisis correctly and to minimize the losses at the airports. For this purpose, in our study, the monthly total number of air passengers was modeled with recurrent neural network (RNN), gated recurrent unit neural network (GRU) and long short-term memory neural network (LSTM) methods. Obtained results were compared with MAE, RMSE, MPE and MAPE criteria. The results show that deep learning algorithms produce successful results in forecasting the number of monthly air passengers. As a result, we observed that the GRU method gives the best forecasting results.