More in this series:
Machine Learning and Analytics in the High Performance Cloud
- Part 1: Ideal Architecture for AI/ML and Analytics
- Part 2: Moving to the Edge: Pushing Compute from the Cloud to the Fringe
- Part 3: Viable Systems Model
- Part 4: Analytics in the Cloud
Analytics seems to be, or had been, the buzzword du jour, but, what exactly is analytics and how is it related to Artificial Intelligence, Business Intelligence, Decision Support, or anything else for that matter?
When addressing topics like this it is always wise to start with a definition. I like what the Wikipedia entry on Analytics says, “Data Analytics is the discovery, interpretation, and communication of meaningful patterns in data.”
Good definition, no? According to our definition, analytics are any sort of activity that extracts meaning from data. Another way of looking at analytics is that an analytic is any sort of measurement or operation that attaches meaningful, descriptive metadata and or semantic information to a piece of data.
For example, let’s look at a lion and let’s see if we can answer the question, “Was this lion born in this area where he was photographed”
What types of analysis can we conduct on our picture of a lion that will yield us any information, metadata that we can attach to the photo and eventually answer our question?
Well, for starters, the lion has a mane, and from there we can conclude he is a biological male as female lions don’t have manes. From the surrounding land we can see that our lion is in a grassland and if we could identify the species of grass we could begin piecing together where this lion lives and what subspecies he is a part of. As it turns out, this lion is a South African lion, Panther leo krugeri, and the photo was taken in Namibia. If we measured this lion we would determine he’s seven feet from nose to the start of his tail and best guess weighs in the neighborhood of four hundred pounds.
Now we’ve measured our lion more ways than a tailor making a custom suit and we can begin to make some inferences/analysis and predictions based upon our measurements and generated metadata. But first, let’s tabulate the metadata to make our lives a bit easier.
Photo has a subject, subject is a lion
- Lion has a genus, species and subspecies is a Panther leo krugeri
- Panther leo krugeri has a habitat is a grassland
- Panther leo krugeri has a size is a measure of 7 feet
- Panther leo krugeri has a mane
- Mane has a color, color is dark
- Mane has a length, length is a long
- Panther leo krugeri has a length, length is a measure of seven feet
- Panther leo krugeri has a gender, gender is a male
- Panther leo krugeri has a weight, weight is 400 pounds (181 kg)
OK, so now we have a set of evidence (descriptors) that we can begin to generate an analysis against.
- The first thing we can assume is that our lion is a mature adult, as manes don’t begin growing until one year of age and darken with age and social status within the pride.
- Given the length of this lion’s mane and the dark color, we can conclude that he is certainly not a juvenile.
- Additionally, from the length and color we can also begin to infer social status. With a mane like this he is in his prime and if not the dominant male in the pride very high ranking.
- Given the color and length of his mane we could estimate that this is a mature adult and not an adolescent male, he’s probably at least seven years old. A dominant male of this maturity will have produced several generations of offspring.
- We can also conclude that he was born outside of the current prides range as male lions have to invade and take over a pride to assert breeding dominance, this structure prevents inbreeding amongst lions.
So, as you can see, some measurements and observations when paired with a little bit of knowledge can generate some very convincing scenarios, conclusions, and yes, analysis. But, how does this fit into AI/ML/DL, all things Big Data, and any other flavor of the moment?
It fits very naturally, if we take a closer look at what we’ve done, we’ve applied some measurements and then mapped some material to those measurements, ex. Panther leo krugeri has a mane, and then mapped additional material until we came to a conclusion(s). The most startling conclusion is that this lion was not born in the area where his photo was taken. Here is where I’d like to make a point. Analytics are used to drive supportable conclusions, not to simply create fancy graphics for others to interpret. The result of an analytic is not a measurement, or picture, it is a supported statement! Our supported statement to the question, “Was this lion born in the area where he was photographed?” is a well-supported, “No!” Can this process be automated and enable data to answer questions? Yes!
With our goal of a supported statement in mind, we see that AI/ML/DL, Big Data methods, semantics, ontologics, and other methods, are all components that assist us in transforming a piece of raw data into a source that can provide a supportable answer to a question. The questions you choose to ask and the methods you employ are unique to your problem, contact us at Nimbix to see how we might be able to help turn your data into something actionable.