抄録
In this paper we present an approach for extracting well-defined and interpretable information granules for classification. The approach, called DC* (Double Clustering by A*) firstly identifies cluster prototypes in the multidimensional data space via the LVQ1 algorithm, then clusters the projections of these prototypes along each dimension by a properly defined search procedure based on the A* strategy. Like DCClass - a previously developed double clustering approach - DC* exploits class information to extract information granules whose granularity level is automatically determined. The added value introduced by DC* is the possibility to extract the minimum number of information granules that give a compact, hence highly interpretable, description of available data. The resulting fuzzy information granules can be directly translated in human-comprehensible fuzzy classification rules. Experimental results on a medical diagnosis problem show the effectiveness of the proposed DC* in comparison to DCClass.
