Journal of Korean Society of Transportation ISSN:1229-1366(Print) 2234-4217(Online) 대한교통학회지

The Air transport industry continues to grow as international trade volume increases due to global economic growth and the emergence of the WTO system. Competition among airlines is intensifying in the global air transport market, and airlines are trying to strengthen their competitiveness through qualitative growth to gain a competitive advantage. In particular, increasing operational efficiency has emerged as a critical factor. Demand forecasting is one of the areas for achieving operational efficiency, and aviation demand forecasting is used as reference material for decision-making by economic players in the aviation industry. Daily air cargo forecasting is essential data for establishing detailed operational plans, such as operational planning and relocation of equipment or personnel from airports or airlines. This paper estimates the daily air cargo prediction model by combining the ARIMA model and the CART (Classification And Regression Tree) and CHAID (Chi-squared Automatic Interection Detection) methodology. And the accuracy of the prediction models was compared and analyzed to identify the accurate daily air cargo prediction model. The prediction model was generated through a total of two steps. In the first step, a weekly prediction model was generated using the number of holidays per week of major trading countries and the weekly ARIMA forecasting result. In the second step, a daily prediction model was generated using weekly prediction results from the first step and the day and holiday information of major trading countries. Based on the forecasting models in this paper, the air cargo for the next 91 days predicted from January, the CART model’s MAPE is 12.20%, the CHAID model’s MAPE is 11.35%, and the ARIMA model’s MAPE is 12.28%. For the air cargo for the next 91 days predicted from July, the CART model’s MAPE is 10.47%, the CHAID model’s MAPE is 11.20%, and the ARIMA model’s MAPE is 11.44%. This study identified which model is suitable for daily air cargo prediction according to the prediction point by comparing and analyzing the accuracy of daily air cargo prediction models. In addition, analyze the cause of the difference in accuracy between the prediction models and suggest implications.