The interpretation of MADMatch results is quite straightforward.
MADMatch results are stored in a csv file with the following headers
Operation | entity1 | entity2 | raw_cost | v1_cost | v2_cost | unmatch_cost | Entity_type | change?
Operation informs about the edit operation reported by a given line:
ns: [entity 1] is matched to [entity2]. The string "--" is used to separate goups of entities
nd: [entity1] is deleted from the first diagram
ni: [entity2] is added in the second diagram
In case of matches (Operation = "ns"), a cost (unmatch_cost) is displayed to inform about how
more expensive it would be to dismiss the proposed match. The higher that cost, the more likely
the match. Note that these costs should be evaluated relatively to the average in a given solution.
In MADMatch, when considering only lexical information, the maximal value which can be paid for
totally dissimilar entities is 100. Thus, one may consider that a match with an unmatch cost below
50 is suspect.
Entity Type informs about the entity type (package, class, method, attribute).
"change?" is a binary value indicating whether the entity has been moved or renamed.