The state of data affairs over the last ten years or so
revolved around big data. Of course, size matters, but big data promises to
morph to monster data as more data sources hit the cloud with more tributaries
like voice, video, IoT, events, and business patterns. So what about all this
parked data? Are we going to keep storing it and bragging about how much cloud
space it consumes? Are you going to make
it cleaner and smarter or just admire it? I would suggest we make data more
intelligent and faster than just figuring out how to catalog and park it, so we
can use it later. Making it faster means treating the data as a database of now,
now of the future. Making data smarter can be tricky, but it is worth it.
Gleaning Data is
Basic Intelligence.
Capturing data of different types and classifying them is
pretty normal. Deciding how long to and where to keep it is essential.
Determining if it is worthy of a long time archiving is doing data a solid.
Knowing some basics about the data source and cost of acquisition and relative
purity is pretty much a given these days. Some data cleansing and organization
will help usage down the road.
Giving Data Meaning
is Average Intelligence
Knowing the data about the data (AKA meta-data) is essential
for interpreting it. The simplest is understating the data’s domain and its
relative relationship to other data (logically or physically). Data
representation and transformation options are pretty essential when combining
with other data. Knowing the key or identifier of groups of related data is
pretty standard. This step is where some of the impurities can be dealt with
before heavy use. First use usually revolves around visualization and reporting
to find actionable insights. This step is turning descriptive data into a
prescription at times.
Granting Data
Representation in Its Context is Very Smart
Most data is gathered and used within one or two base
contexts. One is undoubtedly timing/frequency, and the other is the primary
home of the data. For instance, the entity family it belongs to like product
data. Sophisticated context representation will go beyond an original context
or source to include others that have a neighborhood relationship with the data
grouping/entity. An example would be a product within multiple markets and
channels. This level is where statistical and predictive models enable more
actions to either react or intercept the trends indicated in the data. This
level is turning prescription to prediction to create/place data, event, or
pattern sentinels on processes or the edge to look for prediction completion or
variants.
Grinding Data to a Fine Edge is Smarter
We are interrogating data to learn the need for important
adjustments to goals, rules, or constraints for operating processes that
include humans, software systems, or machines. This level can build a change to
work in a supervised or unsupervised change process. This level starts with
machine learning and extends to deep leading, which peels back layers and
interrogates more data. In extreme cases, the data can be used to support
judgment, reason, and creativity. The worm turns from data-driven to
goal-driven, established by cognitive collaborations with management
principles, guidelines, and guardrails.
Grappling with Data
in Motion Right Now is Brilliance
The pinnacle of smart data is where the data coming in fresh
is used to create the “database of now”. At this level, all of the approaches above
can be applied in a hybrid/complex fashion in a near time/ real-time basis. This
level uses the combined IQ of all the AI and algorithm-driven approaches in a
poly-analytical way that leverages the brainpower combined with fast data. A
dynamic smart parts creation and dynamic assembly line would be a non-combat
example.
Net; Net:
Data: Use it or lose
it, but let the data lead to the learnings that sense, decide, and suggest
responses appropriate to action windows necessary to meet the timing need. If
it is a sub-second focused problem domain, the patterns in the data and
intelligent methods may make the decisions and take action with governance
constraints. If not subs-second focused, let smart notifications or options be
presented to humans supervising the actions. Don't leave all the precious data
parked for the future only.
Building adaptive operational policies (using KEEL Technology) allows systems to interpret complex (dynamic, non-linear, inter-related, multi-dimensional) data sets in real-time and react in real time with what to do, INCLUDING how, how much, where and when to act while considering multiple, sometimes conflicting goals.
ReplyDelete