Statistical learning in the sense of supervised learning refer to a set of techniques aimed at finding a relationship between a set of variables and an outcome. It is now applied to many areas such as finance, biology, marketing and many others. Thank to an extensive area in the field, plenty of techniques are now available. It is then of interest to go through the main ideas which led the development of the most up-to-date techniques (random forests, SVM, neural networks, GAM) starting from the most simple (k-NN, GLM). In this talk we will give a broad overview of the current models in use for predictions and how the data scientist can choose between them in real-world contexts.