摘要 |
<P>PROBLEM TO BE SOLVED: To extract an attribute capable of being additively interpreted without intricateness of positive and negative in analysis, and to perform evaluation based on a relevance degree corresponding to the attribute. <P>SOLUTION: In this evaluation device 1, an input part 10 inputs a plurality of pieces of model-configuring enterprise data shown by an input data attribute, the number of projection attributes, and the number of NMF (Non-Negative Matrix Factorization) attributes. A subspace configuration part 11 configures a subspace composed of a base vector having a magnitude decided by the number of the projection attributes and the number of the input data attributes from the model-configuring enterprise data. A projection score calculation part 12 calculates a value obtained by projection of the model-configuring enterprise data to the subspace as a projection score in each piece of the model-configuring enterprise data, and generates a projection score matrix with the calculated projection score as an element. An NMF calculation part 13 performs NMF on the basis of a transposed matrix of the projection score matrix and the number of the NMF attributes to calculate an influence degree score in each of the NMF attributes capable of being additively interpreted. <P>COPYRIGHT: (C)2008,JPO&INPIT |