Using constraint technology to mine frequent datasets

Constriant-based mining is introduced to efficiently sift the useful itemsets or rules through a large number of mined ones. Two large classes of constraint-based frequent itemsets mining (monotone constraint and succinct constraint) have been investigated. However, the problem of frequent itemsets mining with tough constraint has not been solved just because of the complexity of the constraint. In this paper, we propose a TCA algorithm (tough constraint-based frequent itemsets mining algorithm) which uses the order as the pre-process to solve the problem. The principle of the algorithm is to push the tough constraint deeply inside the candidate generation-and-test approach such as a priori. We also extend it to the multi-constraint case. We conclude that we can improve the speed and efficiency in testing the candidate itemsets through pre-calculating the corresponding conditions under the multi-constraint.