Forecasting the Daily Demand of Air Cargo Using Data Mining with CHAID Approach

Abstract

Since the WTO was launched in 1995, Air cargo demand has risen sharply. It is expected to grow further on the explosive growth of E-commerce and Cross-Border trade in recently. As air cargo demand increases, the importance and needs for the forecasting of air cargo demand is increasing as well. Most of previous researches has been focussed on passenger part. In the case of researches on the forecasting of air cargo demand, the majority of researches are conducted quarterly or yearly forecasting to apply for establishment of mid-/long-term strategies, and an investment plan for an airport. The purpose of this paper is to develope the daily air cargo forecasting model that is able to help players in aviation, airlines, airports, etc., establish detailed operational strategies. In this paper, Chi-squared automatic interaction detection methodology is used to develop the forecasting model. The forecasting model is developed through two steps. At the first step, the weekly volume of air cargo is predicted by using CHAID methodology based on predict value from autoregressive integrated moving average and holiday information. At the second step, the final model which is the daily air cargo demand forecasting model is developed based on the weekly forecasting result from the first step, and holiday information by CHAID method as well. Based on the forecasting model developed in this paper, the daily cargo volumes for the next 56 days are predicted and the forecasting accuracy for each day is 93.9% which is 8.6% point higher than the forecasting accuracy for ARIMA model. It was noted that, unlike the characteristics of general demand forecasts, the high forecasting accuracy is maintained regardless of time lag from the forecasting point. And the result of the forecasting by shifting the forecasting point to 20 point, the forecasting accuracy for each dais is 91.2%, is high as well. The research finding shows the forecast model of this paper is worth to use as a daily forecasting model. It is expected that this paper will help to forecast the daily air cargo demand, and will further be used to forecast daily demand in more diverse area.