Graphics Reference
In-Depth Information
09:58:35 55.44.33.22 GET /ad/advertisement.js 404
...
Sequences can also be found in many other kinds of data. One good example
is the creation of a network of doctors based on patient visits by creating
links for multiple doctors who bill for the same patient around the same
time (
http://bit.ly/1bgyHuk
)
. Strong links between a pair of doctors implies
a strong relationship (such as referrals). What's fascinating here is that raw
transaction data of patient visits has been turned into a graph, which then
reveals new, valuable information about connections between doctors.
Another example is patent citations. Each patent references prior patents.
The various references can be collected to gain insights such as which
patents are referenced the most.
Unstructured Data (for Example, Tweets)
Unstructureddatacanalsobeprocessedtoextractnodesandlinks.Ameans
to identify nodes and identify links is required. For example, tweets are
short, 140-character messages publicly broadcast on Twitter. Tweets are a
rich data source from which you can mine different kinds of nodes and
links by looking for co-occurrence of hash tags (that is, user-defined topics),
usernames, or stock symbols within tweets, and you can extract these to
form graphs. This approach is similar to the transaction approach used with
e-mail analysis described earlier. Sample raw data may look like this (for
example, via
tweetarchivist.com
):
UserName, Time, Tweet
Benzinga, 01/15/2014, Is #Wendys Success at #McDonalds
Expense?
$MCD $WEN http://t.co/OibzrKFiVB
SeekingAlpha, 01/15/2014, 2 Dividend Machines I
Purchased Last Week
http://t.co/hMcX5rvSxH $TGT $KO $MCD
wallstCS, 01/15/2014, RT @Jacqui_WSCS: #Starbucks
Catches
"McDonald's Syndrome" and Gets a #Stock Downgrade
http://t.co/elwMdFbcQ4 via @wallstCS $SBUX $MCD
...
