The value of analytical queries on Social Networks

We study analytical queries for two models of Social Networks. Firstly, a datawarehouse which can be analyzed along some OLAP schema and possible dimensions, secondly a model of streaming data which need to be transformed before they are analyzed. In the first case, the linear influence model is generalized for each possible dimensions and provides densities of influence types. In the second case, we need to approximate the analytical queries by sampling the data along some specific distributions. For a model of analytical queries on graphs and their approximations, we give examples of approximable and non approximable queries. We introduce a measure to quantify the information provided by various analyses using both the Entropy of the answer to the query and of the influence types. We illustrate the approach on Facebook and Twitter data.