Forecasting Inbound Logistics for Express Cargo Transportation: A Case Study of Turkey

Abstract

Within the domain of inbound logistics, the express air cargo transportation sector has become an essential component of global trade. As the disparity between actual demand and forecasted demand in express cargo transportation widens, the potential for resource wastage correspondingly increases due to the unpredictability of volume and weight. Therefore, this study aims to forecast the daily quantity and weight of incoming cargo, categorized by type, within the context of inbound logistics. Utilizing a case study of express cargo transportation in Turkey, we employ both Long Short-Term Memory (LSTM) and Seasonal Autoregressive Integrated Moving Average (SARIMA) models to compare the forecasting performance of the LSTM approach. In the LSTM model, the maximum epoch, batch size, number of neurons and optimizer parameters are adjusted using grid search to reduce the prediction error. This forecasting capability enables businesses to better prepare for sudden fluctuations in incoming shipments and provides a methodological and analytical framework that influences daily operations. Additionally, we seek to contribute to the existing literature on operational planning by developing a model capable of generating daily forecasts, as opposed to traditional forecasting models that operate on different temporal scales. The numerical results indicate that the improved LSTM model outperforms the SARIMA model for all data sets.