MODELLING AUSTRALIA’S EXPORT AIR CARGO DEMAND USING AN ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM APPROACH

Abstract

This paper proposes an adaptive neuro-fuzzy inference system (ANFIS) model for predicting Australia's annual export cargo demand, as measured by enplaned tonnage. The study used annual data for the period 1993 to 2016. The data was divided into two discrete data sets. The first was used to train the ANFIS, whilst the second was used for model estimation. The data was normalized to increase the ANFIS model's training performance. Sugeno fuzzy rules were used in the ANFIS structure. Gaussian membership function and linear membership functions were developed to optimize the model’s performance. The hybrid learning algorithm and the subtractive clustering partition method were used to generate the optimum ANFIS model. In the computational analysis, the predictive capability of the ANFIS was examined for the following ranges of clustering parameters: range of influence (ROI), squash factor (SF), accept ratio (AR), and reject ratio (RR). The results indicated that the ROI, SF, AR and RR were obtained to be 0.50, 1.25, 0.50 and 0.15, respectively, for the optimum fuzzy inference system (FIS) structure. The mean absolute percentage error (MAPE) for the out of sample testing dataset was 3.42%. The actual R2 value of the final ANFIS model was 0.9857%, demonstrating that the model has a high predictive capability.