Einsatz prädiktiver Analytik und maschinellen Lernens in der Supply-Chain-Nachfrageprognose

Abstract

The global economy, which is characterized by constant change, makes it increasingly difficult for companies to create accurate demand forecasts. To counteract this problem dynamic forecasting methods are required that do not only deal with simple demand signals such as trends or seasonality. Two methods for forecasting that have received a lot of attention in recent years are predictive analytics and machine learning, which are already increasingly being used in supply chain management. However, introducing these methods into companies is a complex process that involves several steps. This is why it is necessary to create and follow a kind of roadmap for implementing predictive analytics and machine learning used in supply chain demand forecasting. The first chapter of this thesis deals with the basics of predictive analytics in supply chain demand forecasting. In addition to the definition of supply chain demand forecasting, its objectives, basic concepts, and methods are discussed. Predictive analytics is then described in more detail before explaining how predictive analytics can be used in supply chain demand forecasting. The following chapter discusses machine learning, highlighting the difference between machine learning and predictive analytics. Furthermore, the suitability of machine learning algorithms for supply chain demand forecasting is considered before an implementation recommendation for predictive analytics and machine learning in supply chain demand forecasting for an electrical industry company is drawn up in the concluding chapter. In the last chapter of this thesis, an implementation recommendation for an electrical industry company from Upper Austria is drawn up, considering the industry standards for projects with predictive analytics (CRISP-DM) and machine learning (CRISP-ML (Q)). This recommendation provides a guideline for the company on which steps should be followed in which order in a project dealing with the implementation of predictive analytics and machine learning and what should be taken into account in the respective project phases. This implementation guide consists of six phases that consider the characteristics of both predictive analytics and machine learning. This recommendation enables the company to minimize the risks associated with the possible introduction of predictive analytics and machine learning in supply chain demand forecasting. Furthermore, unnecessary work steps and thus valuable working hours can be avoided by adhering to the implementation recommendation, as it focuses on the essential implementation measures and thus avoids an unstructured implementation.

Date of Award	2024
Original language	German (Austria)
Supervisor	Gerald Schönwetter (Supervisor)