A fuzzy structure similarity algorithm for attributed generalized trees

In this paper, our vertex-attributed edge-attributed generalized tree structure proposed earlier is augmented using fuzzy attributes. Labels of vertices represent objects, while edge labels express fuzzy attributes. Edge weights represent the (percentage-)relative importance of fuzzy attributes, a kind of pragmatic information. The generalized trees are uniformly represented and interchanged using a fuzzy extension of Weighted Object Oriented RuleML. In the process of matching two generalized trees, a set of membership degrees related to the linguistic terms of fuzzy sets is assigned to each vertex using fuzzification of the numeric data of vertex labels. The fuzzy similarity of membership degrees related to each pair of corresponding vertex labels is computed, and the obtained fuzzy similarity value is considered in the structure similarity process. It is shown that this approach outperforms our earlier generalized tree similarity approach that considers exact string matching for computing the similarity of vertex labels. The use of our approach is demonstrated for life-insurance application underwriting.