Generalized association rule mining with constraints

Generalized association rule extraction is a powerful tool to discover a high level view of the interesting patterns hidden in the analyzed data. However, since the patterns are extracted at any level of abstraction, the mined rule set may be too large to be effectively exploited in the decision making process. Thus, to discover valuable and interesting knowledge a post-processing step is usually required. This paper presents the CoGAR framework to efficiently support constrained generalized association rule mining. The generalization process of CoGAR exploits a (user-provided) multiple-taxonomy to drive an opportunistic itemset generalization process, which prevents discarding relevant but infrequent knowledge by aggregating features at different granularity levels. Besides the traditional support and confidence constraints, two further constraints are enforced: (i) schema constraints and (ii) the opportunistic confidence constraint. Schema constraints allow the analyst to specify the structure of the patterns of interest and drive the itemset mining phase. The opportunistic confidence constraint, a new constraint proposed in this paper, allows us to discriminate between significant and redundant rules by analyzing similar rules belonging to different abstraction levels. This constraint is enforced during the rule generation step. Experiments performed on real datasets collected in two different application domains show the effectiveness and the efficiency of the proposed framework in mining constrained generalized association rules.