@inproceedings{TrahaschPrinzbach2018,
  author    = {Stephan Trahasch and J{\"u}rgen Prinzbach},
  title     = {Predictive Analytics in Utility Vehicle Maintenance},
  series = {DATA ANALYTICS 2018 : The Seventh International Conference on Data Analytics},
  isbn      = {978-1-61208-681-1},
  issn      = {2308-4464},
  pages     = {97 -- 102},
  year      = {2018},
  abstract  = {In public transportation, the motor pool often consists of various different vehicles bought over a duration of many years. Sometimes, they even differ within one batch bought at the same time. This poses a considerable challenge in the storage and allocation of spare parts, especially in the event of damage to a vehicle. Correctly assigning these parts before the vehicle reaches the workshop could significantly reduce both the downtime and, therefore, the actual costs for companies. In order to achieve this, the current software uses a simple probability calculation. To improve the performance, the data of specific companies was analysed, preprocessed and used with several modelling techniques to classify and, therefore, predict the spare parts to be used in the event of a faulty vehicle. We summarize our experience running through the steps of the Cross Industry Standard Process for Data Mining and compare the performance to the previously used probability. Gradient Boosting Trees turned out to be the best modeling technique for this special case.},
  language  = {en}
}