Market prediction for the automotive after-sales market

Time series are used in many different industries and their specific characteristics make them particularly difficult to accurately forecast. My research examines different variations of artificial intelligence (in particular: recurrent neural networks) and their impact on the performance of time series forecasting. AI algorithms are basically black box solutions. In order to generate user confidence in the resulting forecasts, among other things, it is essential to elucidate this black box.

As part of an industrial doctorate at a German automotive manufacturer, I am investigating the short and medium-term forecasts of revenue planning in after sales. My research juxtaposes variations of forecasting models in different international contexts with the aim of providing a robust statement on which model variation is most adequate. After sales is a highly complex area of business that is influenced by numerous factors, which increase the relevance of determining input variables that have a positive or negative effect on business success.

The findings of this thesis close the research gap of an explainable and comprehensible time series forecast in an international environment based on the selection of the most country-specific robust model. An additional practical implication to be emphasized is the more accurate revenue steering in terms of both content and timing.