Would it be from social networks (Facebook, Twitter for instance), or from their own data (such as email exchanges), organizations have access to important amounts of data. This data may be of specific nature, namely, instead of a description of each observation by characteristics, relations between observations may be known. Studying these relations involves specific techniques. The nature of the data «relations between objects » makes sampling a difficult task. For that reason, it is usually better to work on the whole dataset, which is challenging from the algorithmic point of view.