Model driven engineering for data miners simulation

Abstract

Decision making processes in some contexts, such as construction of infrastructures, require the design of prospective experiments. These experiments are designed to test hypothesis raised in the design phase. Following an adaptive methodology, these experiments will feed new iterations. In many cases, the development of these experiments is so costly that they must be performed using simulators. In this paper, an approach for developing data mining simulators based on Model Driven Engineering is presented. These miners can be fed from data stores (i.e. multidimensional data stores) and may generate new data stores with the results they provide. This approach is intended to reduce the uncertainty at decision making as well as to speed up the iterative process.