One advanced feature nowdays is to Display a Network in a more Visual Form. This is often applied to Social Networks as you have people communicating and maybe influencing each other, but even for a more Traditional Scenario like a Telephone Network.
Oracle Provides a Standard Data Model for the Telecommunication Industry to keep e.g. all the CR (Call Record) Data. One Way to use this CR Data is to Visualize a Network based upon the Call Records and Identify People that might churn and check which affect this might have on other Customers. The Oracle Product of OCDM is further described here. Additionally a Tutorial is also provided here.
The Churn calculation is performed by another Algorithm of the Standard Data Model, but now this Information can be visually included within the Network Analysis.
Thus, we want to display the following Information:
- If a Customer is subject to Churning
- Call Volume between Network Participants
- Revenue Generated by each Network Participant
We supply the above Information using to Kind of Data Sets:
Descriptive Attributes of the Network Participants (Nodes) including Revenue Information (From the Billing System) and Churn Information (Calculated by another Data Mining Algorithm):
And the Information about the Call Volumn between the Network Participants/ Nodes:
To do this we can Export the very Structured Data from the Oracle Communication Data Model and load use a Open Source Software e.g. Cytoscape. (in the following Examples 2.6.2 has been used).
We can immediately see that the Network is visualized and gives an Impression of the Information. If we would have just the Call Information in a tabular format like above this would have been incomprehensibly difficult for Humans to Understand and interpret.
We have many Configuration Options available for this Diagram which we can use to map and Visualize our Information. E.g. we can use the Call Volume expressed as TOTAL_CALL_CNT (this has been Aggregated from the Detail Level Records beforehand) as the Line width, the Churn Indicator as the Node Color and the Node Size is the Revenue (as Revenue Band, which has also been Aggregated beforehand).
When Scrolling in one can see that the Line Width is denoting the Call Volume between the individual Nodes and the Color Red/Green indicates the Nodes with a Churn Risk:
We can now use the VizMapper to further enhance the readability of the Network.
One can also just select the nodes for further Analysis and see exactly which kind of nodes belong to a certain Part of the Network. In the below example the top right corner has been highlighted and to have more clear Vizualisation the Style has been changed to Universe in the VizMapper Property Section.
Using a Open Source Software like Cytoscape and the Oracle Communication Data Model only relative moderate Data Preparation and Visualization Effort is required to perform the Network Analysis.