Pre-selection of Suitable Regression Methods for the Determination of Interactions and Forecasts in Global Production Networks

Abstract

The locations of many manufacturing companies are distributed globally. This has led to the development of historically grown global production networks whose structure is often very complex, not transparent and influenced by many factors. The high number, as well as the volatility of the influencing factors and dependencies in the network additionally, complicate the network configuration. As a result, adaptation needs and optimization possibilities are recognized too late or not at all. In order to enable early recognition of saving potentials, active monitoring and analysis of changes and dependencies of the influencing factors on the production network is needed. The necessary consideration of a multitude of influencing factors requires further tools to be manageable by the network planner. Therefore, databased methods can be used as support for the forecast and the determination of dependencies of influencing factors. In other research fields, regression analysis is an established method for a databased analysis. This paper focuses on the use of regression analysis in global production networks. It is essential for an accurate analysis, to choose the right regression method out of the many different types in existence. A systematic literature review is conducted to establish an overview of regression methods used in other research fields. A search strategy is developed and implemented and the key findings of the literature review are derived and evaluated. In the second step, a new approach for the pre-selection of suitable regression methods for the determination of interactions and forecasts in global production networks is proposed.