Graphics Reference
In-Depth Information
Link Records (for Example, Network Logs)
Sometimes only links are identified in a data set. One example is network
log files. Although log files may seem arcane, they contain a wealth of
interesting information—for example, from where people are connecting
into a corporate network, when and where big files are transferred out,
patterns of regular activity (such as network backup), and patterns of
irregular activity (such as hackers attempting to break in).
NetworklogfilesmayindicateasourcecomputerIPaddressanddestination
computer IP address in each line, identifying a link. You can extract nodes
by compiling a list of all unique IP addresses in the source and destination
IP addresses, as shown here:
Timestamp, Source, Destination, Etc
2/25/2014, 9.8.2.3, 128.2.9.87
2/25/2014, 7.6.9.5, 128.2.19.45
2/25/2014, 7.6.9.5, 128.2.9.87
...
For an example, see the
Network_Logs
data in the Supplementary
Material on this topic's companion website. This is a small sample Excel
spreadsheet showing anonymous network data from an intrusion system,
and it shows techniques for summarizing the link data using pivot tables.
Real-world network systems can generate tremendous amounts of data that
require a programmatic approach to handle the data.
Transaction Records (for Example, Purchases, E-mails)
By looking at the items that co-exist in a transaction, you can construct
a graph. Nodes are the items, and links are the co-occurrence of items
within any transaction. Examples ofthistypeofgraph include awide variety
of social networking (including e-mail data, as discussed earlier in this
chapter), as well as multiple authors of documents such as topics, news
stories, or reports.
Ever wonder how websites recommend additional products you may be
interested in? If you're looking at the topic
Calvin and Hobbes Lazy Sunday
Afternoon
(Kansas City, Mo.: Andrews-McNeel Publishing, 1989) and see
four other recommended topics, these additional topics could be the result
of a market basket analysis.
