Why use link prediction?

Link prediction consists of guessing which unknown connections between existing nodes in the graph are more likely to be real, now or in the future. With a proper formulation, this can be turned into a machine learning problem. But before trying to build a model for link prediction, as for any data science task, we must start with a deep understanding of the problem. That will be our goal in this section. First, by using the context of dynamic graphs, we will define what exactly is hidden behind link prediction. We will also review some applications, from marketing to science.