Towards predictive analytics in internal logistics – An approach for the data-driven determination of key performance indicators

Data-driven methods can leverage opportunities to optimize internal logistics systems. Performance metrics can be gathered by Machine Learning. However, missing information is a significant obstacle. Therefore, additional data is necessary. This article presents a procedure model for process analysis, from database preprocessing to gathering process insights, focusing on predictive analytics. A control theory-driven approach categorizes the data, being the guideline for the procedure model. Assistance is provided in selecting data processing methods and obtaining valuable insights. A case study validates the procedure model with a performance analysis by predicting breakdown occurrence, comparing k-nearest neighbors, decision tree algorithms and neural networks.