Bus detection equates to finding the highly integrative nodes in a network (graph). The terminology
bus originates from the bus component often found in electronic hardware. Most other components
plug into or communicate via the bus. The bus detection algorithms are grouped into the
Bus nodes often display high node degrees (number of incoming and outgoing edges) or other measures
of centrality. Currently the only algorithm that is implemented is the
detection. It utilizes the degree distribution of nodes and distinguishes bus
nodes by a factor gamma with respect to the median of node degrees within a graph (or graph
Detecting buses does not change any values or properties within the graph and merely returns the detection results. This is on purpose, as you might provide any groups of nodes that aren't necessarily siblings in any way.
Let's use the Ford Climate Control System dataset that is a graph that models the system architecture of a climate control system you would find in your car. We can use the following snippet to return the names of the nodes that would be considered bus nodes and those that are not:
Where we can see that we detect three nodes as the bus nodes for
It is up to the user to interpret the results of this analysis! (or any analysis, really)
Suppose we want to incorporate these results in our
Graph, we could then do
We add a parent
Node here named
"system" to indicate the system boundary.
This is intentional and even required! The
is_bus property namely
indicates that a Node is a bus within or for the parent (sub-)system that we just defined.
For instance, you could have a complete hierarchy of nodes and have a local bus node that isn't highly integrative for your the complete system (all nodes), but does fulfill that role in a sub-system located deeper in your hierarchy.