Optimizing Air Freight Forecasting: The Integration of Machine Learning and Empirical Mode Decomposition

Accurately forecasting freight volumes in airports is crucial for optimizing logistics operations and enhancing supply chain efficiency. This study focuses on predicting the freight volume transported through Australian airports (measured in kilograms) using advanced machine learning techniques. To improve forecasting accuracy, the Empirical Mode Decomposition (EMD) algorithm was applied to decompose the time series into sub-components (Intrinsic Mode Functions - IMFs) and a residual component. Each subcomponent was then individually forecasted using machine learning models, and the final predictions were obtained by aggregating the results. The study employed five machine learning techniques: Support Vector Machine (SVM), Artificial Neural Networks (ANN), Classification and Regression Trees (CART), Gaussian Process Regression (GPR), and Generalized Additive Models (GAM). To enhance model performance, the hyperparameters were optimized using Bayesian optimization. Additionally, a comparative analysis was conducted by reapplying the models to the original, non-decomposed time series to evaluate the impact of EMD on forecasting accuracy. The findings indicate that models incorporating EMD significantly outperform those applied directly to the raw time series, achieving lower error metrics and improved predictive accuracy.