Automatically Generate a Flat Mining Table with Dataconda

Classification and regression algorithms require a flat mining table as input, which in most cases is built manually by summarizing relational data into propositional features. This task is not only time consuming, but it also inhibits the discovery of new knowledge, because a small portion of the possible features will be built. Dataconda, a software program available on www.dataconda.net, makes this task automatic, thereby facilitating new knowledge to emerge. The user selects a target variable from any table of a relational database, and Dataconda builds and tests a large number of predictors by aggregating information from the other tables. For example, Dataconda may find that the best predictor for "customer loyalty" is the amount of money spent by the customer in cheap products, even if the user has not built any such feature. This demo will illustrate how to use Dataconda in a tutorial database and in a real-world database.