Graphs are increasingly used to describe interactions between entities. They are based on simple formalism that nevertheless allows modelling of complex systems such as industrial ecosystems. Thus, a knowledge graph can be built from traditional economic variables, but also from new alternatives variables from open-source’s data and big data. In this article, we review some graph learning methods and discusses latest advances in this field. Machine and deep graph learning method learn embeddings for nodes/edges in a graph to perform many tasks, such as link prediction, clustering, and nodes classification. Originality of this talk is the application on graph learning’s methods to analyze and support industrial resilience. Indeed, by learning knowledge graphs which represents an industrial ecosystem, we could help for a more resilient and ecological production. In this survey, we discuss on many advantages of graph learning models and their limits. This review shows latest advances and leads to a discussion of prospects for future research.