Airline Passenger Forecasting using ARIMA and Artificial Neural Networks Approaches

Demand uncertainty has been increasing as a result of the rising trend of using airplanes as a transportation mode option in Indonesia over the years. This condition results in the need for the ability to accommodate the rise for airline companies to withstand within the industry. The forecast accuracy highly determines strategy formulation. Thus, accurate forecasting models are crucially needed. In this study, neural network is proposed to create the best-fitted model to predict future values as a non-traditional method that has already been tested to result in accurate predictions. As a comparison with the traditional model, Autoregressive Integrated Moving Average (ARIMA) model is applied. This study used monthly passenger data from Indonesian airlines, focused on Jakarta-Yogyakarta (CGK-JOG) and Jakarta-Singapore (CGK-SIN) routes, which are the representatives of the most profitable route for both domestic and international flight. Mean Absolute Percentage Error (MAPE) of both methods were then compared and forecasted future demand for the next 12 months were calculated. In both routes, neural network produced better value than ARIMA with MAPE of 1.29 for the CGK-JOG route and 1.66 for the CGK-SIN route.