Comparing statistical and machine learning methods for time series forecasting in data-driven logistics -- A simulation study