Application of forecasting methods to passenger traffic volume and load factor prediction: Wizz Air airlines case study

For passenger traffic forecasting on the testing dataset, the best performance is achieved by the additive Winters model, followed closely by the naïve methods. In contrast, the weakest results are observed for the trend model. In the training phase, the lowest errors were obtained for My_ARIMA, Auto_ARIMA, and the additive Winters model, while the naïve approach performed worst. When averaging results across training and testing datasets, the additive Winters model demonstrates the overall lowest MAPE, indicating strong and stable forecasting performance. Forecasting results for the load factor differ slightly from those for passenger volumes. Overall, all models achieve improved accuracy in load factor prediction, suggesting that load factor exhibits more regular and learnable patterns over time. The most accurate forecasts for load factor are produced by Auto_ARIMA and the additive Winters model. Conversely, the trend model yields the highest forecasting errors.