Kuwertz, AchimAchimKuwertzMoll, MaximilianMaximilianMollSander, JenniferJenniferSanderPickl, StefanStefanPickl2022-03-142022-03-142020https://publica.fraunhofer.de/handle/publica/40804210.1007/978-3-030-27928-8_78Given the importance of early warning in crisis prevention this paper discusses both knowledge-based and data-driven approaches. Traditional knowledge-based methods are often of limited suitability for use in crisis prevention and management, since they typically use a model which has been designed in advance. Novel data-driven Artificial Intelligence (AI) methods such as Deep Learning demonstrate promising skills to learn implicitly from data alone, but require significant computing capacities and a large amount of annotated, high-quality training data. This paper addresses research results on concepts and methods that may serve as building blocks for realizing a decision support tool based on hybrid AI methods, which combine knowledge-based and data-driven methods in a dynamic way and provide an adaptable solution to mitigate the downsides of each individual approach.enearly warningexpert knowledge modelsdeep learning004670conference paper