Combining Machine Learning and Active Objects for Parallel Data Mining

Nowadays, the necessity and usefulness of the field of Data Mining (DM) and Knowledge Discovery in Databases (KDD) are largely established by both the scientific and industrial communities; and a number of real applications have already been developed in domains ranging from space data to financial analysis [1]. However, the need for scaling up DM algorithms is a natural requirement of the more and more huge size of real world Databases (DB). In this work, we are interested in the task of classification based on rule induction technique. The paper presents the system IFORD (Induction of First Order Rules from Databases), designed for learning first order rules from relational DB. The serial version of IFORD is explained in section 2. The next section explains how active objects are used to allow the parallel execution of the mining process. We conclude the paper by presenting our first experiments and future directions.