This article presents a multi-indicator approach to forecasting the monthly workload of a military road freight transport system in support of operational planning. The empirical basis of the study consisted of real-world operational data from 2020 to 2025, aggregated into regular monthly time series. Four complementary workload indicators were analysed: the number of transport tasks, the number of vehicles assigned to task execution, the mass of transported cargo, and transport work expressed in tonne-kilometres. The research procedure comprised data preprocessing, indicator construction, seasonality analysis, time-series decomposition, comparison of classical forecasting models, and assessment of forecast uncertainty using prediction intervals. The forecasting models considered included the naive model, the moving-average model, Brown’s and Holt’s exponential smoothing models, ETS, and ARIMA. Model performance was evaluated using a rolling-origin validation procedure with an expanding training window, based on MAE, RMSE, MAPE, MASE, and Bias metrics. The results showed that the recommended model depends on the forecasted workload dimension: Brown’s model performed best for the number of transport tasks, ETS for the number of vehicles and transport work, whereas the 12-month moving-average model was most effective for transported cargo mass. All recommended models achieved MASE values below 1, indicating improved predictive performance compared with the naive benchmark. The study demonstrated that point forecasts supplemented with 80% and 95% prediction intervals can support monthly planning of fleet resources, transport capacity reserves, and future workload levels. Although the empirical analysis concerns a military transport system operating under peacetime conditions, the proposed framework may be adapted to support monthly workload forecasting and operational planning in other freight transport systems.
Konwerski et al. (Fri,) studied this question.