Databases Reference
In-Depth Information
5.3 Discovery Using Data at Rest
Statisticians and video editors are the third set of team members in a sports-
cast. These people work with a lot of historical data and constantly work on the
statistics to compare the current game with past ones. They also capture and
edit replays by focusing on the game from a start time to an end time and using
multiple camera angles to show the action from different viewpoints.
In the world of Big Data Analytics, statistical models perform structured data
analytics, while a number of accelerators have been built for unstructured analy-
sis. For example, query tools in Big Insights can be used to focus on data from
one point in time to another and correlate a number of sources, much like the way
an editor brings together multiple camera moments in sports television.
Business Intelligence has been focused primarily on structured data. MDM
provides the ability to match structured customer or product data to bring a
single view of customer or product. Data Warehouses provide the ability to store
ingested data and build different data representations for the purpose of reporting
or predictive modeling.
The presence of unstructured data brings in a set of data scientists who work
with a mix of unstructured and structured data to determine insights. These
insights are offered to the orchestration layer and could also be obtained at the
command of the orchestration layer focused on speciic entities (for example,
detailed analytics about the hour before a major network outage, focusing on
network entities in the vicinity of the outage).
The discovery layer provides much-needed relection on the historical data.
By nature, discovery is slower than the conversation and orchestration layers.
However, the results of discovery can be integrated in the analytics engine with
more recent data and used to interpret or augment the conversation. Much as a
commentator may use statistics to make a point in the middle of a narration, the
discovery layer in the Big Data architecture provides much-needed context and
behavioral prediction to a sales conversation in order to provide an offer.
Computers have traditionally been assigned to numeric computing. As a
result, structured data analytics using statistical models is a mature topic. We are
now seeing the emergence of a couple of unstructured data analytics techniques,
as I summarized earlier in the topic (in Section 4.2). Let us turn our attention now
to how we combine inferences drawn from structured and unstructured analysis
and how we use them together for an overall ranking.
Search WWH ::
Custom Search