Today's rule mining algorithms all use greedy approaches to generate rules representing the knowledge hidden in vast amounts of data. When using a greedy approach, systems cannot guarantee that optimal rules are found. On the other hand, exhaustive searhc algorithms find optimal rules. But due to the vast search spaces. exhaustive search algorithms are in most cases impractically slow. This paper presents the A-like rule mining algorithm DA-2, Similarly to exhaustive search algorithms, DA-2 also finds optimal rules. Its running time, however, is just slightly longer than the running time of greedy algorithms.
展开▼
