After reading this chapter, you should have a clearer understanding of what link prediction means and how it can be used to tackle many graph-related questions. You should also know which kind of metric can be used to predict how likely a link is to appear between two nodes in the future. Finally, you have built from scratch a link prediction problem, understanding how it is different from a classical data science problem, and have learned how to successfully build a predictive model to foresee new relationships in a graph.
Until now, we have learned how to build features based on the fact that our data forms a graph structure. It is an important step to understand the graph structure and the prediction power of these features. However, modern machine learning techniques tend to avoid the feature engineering steps where algorithms automatically learn features called embedding. Applying this technique to graphs is the topic we will cover in the following chapter.
