| 摘要 |
PROBLEM TO BE SOLVED: To efficiently extract a frequent part graph commonly included in a plurality of objects with considering superordinate conceptualization into consideration. SOLUTION: This frequent pattern extracting device extracting a set of frequent part graphs included in the plurality of objects comprises a conceptual hierarchy database storing correspondence between a plurality of kinds of components and their upper conceptual elements, an extraction candidate part graph creation section creating a set of extraction candidate part graphs from previously extracted frequent part graph, a weighted support degree calculation section calculating the weighted support degree of the extraction candidate part graph, an upper concept candidate selection section selecting a second extraction candidate part graph as the upper concept pattern of the first extraction candidate part graph, and a frequent part graph addition section adding the second extraction candidate part graph to the set of the frequent part graphs on condition that the weighted support degree of the first extraction candidate part graph is not the same as that of the second extraction candidate part graph. COPYRIGHT: (C)2005,JPO&NCIPI
|
