Conclusion - Data Warehouse Systems: Design and Implementation - page 577

directions), the summarized graph will contain a cycle over the male node,
annotated with a '2'. If there was just one connection between a woman and
a man, there will be an edge between nodes male and female annotated with
a'1'.
Note that from a modeling point of view, there is no agreement upon a
conceptual model for graph databases. To fill this gap, in [ 62 ] the authors
introduce a conceptual model for graph databases, oriented to allow analysts
to perform data analysis over graphs not only in an OLAP style but also
using more sophisticated analysis like data mining, for instance. For this,
as usual, measures and dimensions must be defined. The authors identified
two kinds of measures: informational measures, which are calculated from
the attributes of the edges and nodes, and structural measures, which result
from the algorithms performed on the structural properties of the graph. A
structural measure could be, for instance, a subgraph containing the shortest
path between two nodes or a numerical value computing the length of the
path. Graph evolution is central to this model. This would allow, for example,
to study the evolution of the shortest path between users and products or
the shortest path between two members of a group in a social network.