Umsatzvorhersagen in der Stahlindustrie: Integration des menschlichen Wissens im Predictive Forecasting am Beispiel der voestalpine Steel Division

Abstract

A world without conflicts, events or disasters is unthinkable in this day and age. Shipping congestion, coronavirus and global conflicts are ongoing issues. International companies are therefore faced with the challenge of continuously adapting to changing conditions and have to react faster than ever. Against this backdrop, forecasts must be adapted with increasing agility to be able to respond to changes with immediate measures and target adjustments. A key factor in the general forecasting process is human knowledge. Not only do people have knowledge that plays an important role in the forecasting process, but their opinions can also contribute to distorting the forecasting process. Data-supported forecasting models can be used to avoid such challenges. To effectively integrate the human factor into the model, especially as a data source for one-off events or unpredictable situations, it is crucial to develop methods for optimally integrating intuitive assessments of the current steel market. To answer these research questions, the subject areas are first elaborated theoretically. In the first step, the different forms of business analytics are considered and explained and how a data mining model can best be introduced. Subsequently, success factors and challenges during implementation are discussed based on various case studies. It is shown how the human factor can be integrated into the introduction of predictive solutions. An online questionnaire was developed to find out how the gut feeling about the current market situation on the steel market can be most effectively integrated into predictive models and to identify which data sources sellers see as important indicators for price turning points. The data obtained was evaluated using frequency counts and qualitative content analysis to conclude the integration of the human factor for the voestalpine Steel Division. Furthermore, possible new data sources for the predictive forecasting model are presented. This work makes it clear that, according to the current state of knowledge, human knowledge is an indispensable component for the functioning of predictive forecasting models. In the future, precise forecasts will require both the implementation of such solutions and the essential knowledge of salespeople for unforeseeable events. In addition, the results of the predictive forecast must be checked for plausibility and the data must be interpreted correctly. New data sources could also be identified for possible integration into the future predictive model and conclusions could be drawn as to which data sources the sales experts see as important indicators for price turning points in the steel industry.

Date of Award	2024
Original language	German (Austria)
Supervisor	Christa Hangl (Supervisor)