Fast methods with magic sampling for knowledge discovery in deductive databases with large deduction results

The ability of the deductive database to handle recursive queries is one of its most useful features. It opens up new possibilities for users to view and analyze data. This ability to handle recursive queries, however, still cannot be fully utilized because when recursive rules are involved, the amount of deduced facts can become very large, making it difficult and sometimes impossible to store, view or analyze the query results. In order to overcome this problem, we have proposed the DSK method and the DSK(S) method to discover characteristic rules from large amount of deduction results without having to store all of them. In this paper, we propose two new methods, the DSK(T) method and the DSK(ST) method which are faster than the DSK method and the DSK(S) method respectively. In addition, we propose a new sampling method called magic sampling, which is used by the two methods to achieve the improvement in speed. Magic sampling works when linear recursive rules are involved and the magic set algorithm is used for deduction.