Finding "early" indicators of UML class diagrams understandability and modifiability

Given the relevant role that models obtained in the early stages play in the development of OO systems, in the recent years special attention has been paid to the quality of such models. Adhering to this fact, the main objective of this work is to obtain "early" indicators of UML class diagrams understandability and modifiability. These indicators will allow OO designers to improve the quality of the diagrams they model and hence contribute improving the quality of the OO systems, which are finally delivered. The empirical data were obtained through a controlled experiment and its replication we carried out for obtaining prediction models of the Understandability and Modifiability Time of UML class diagrams based on a set of metrics previously defined for UML class diagrams structural complexity and size. The obtained results, reveal that the metrics that count the number of methods (NM), the number of attributes (NA), the number of generalizations (NGen), the number of dependencies (NDEP), the maximum depth of the generalization hierarchies (MaxDIT) and the maximum height of the aggregation hierarchies (MaxHAgg) could influence the effort needed to maintain UML class diagrams.