Representative End User Clients
Representative Automation Clients
Representative Software Clients
Note: This is the first blog of a two-part series examining some of the differences among advanced analytics techniques. This post discusses some of the differences between cognitive analytics and single-layer algorithm analysis, discussing the example of asset failure prediction as part of an overall asset performance management (APM) strategy.
What Do Cognitive Analytics Try to Do?
Cognitive analytics methods are designed to replicate human thought processes in computers and machines. They are often marketed using terms such as deep learning, deep machine learning, neural networks, neural nets, and artificial intelligence, though in industrial applicat
ions they are often combined with several other analytical methods. Cognitive analytics are experiencing a resurgence, particularly for business use, as the historic challenges of computing them can now be managed (calculating and accurately training these many-layered analytics had been massively difficult).
Using machine learning, cognitive analytics can find representations in data without being told what to look for. So, they tend to work well with unstructured data. This capability is the basis for one type of predictive machine learning used for asset failure and predictive maintenance solutions.
In contrast, statistical and single-layer reasoning algorithms, like decision trees, draw conclusions. They are very effective at doing so and can be applied via machine learning, too, for asset failure and predictive maintenance solutions. These methods are ideal for discovering hidden but known insights by filtering out noisy, irrelevant data.
A Simple Example
A simplistic example is helpful for demonstrating the difference between the two methods. An experienced farmer understands that something sharp shows up in his haystacks on occasion, causing big issues with people who buy the hay. Using single-layer analysis, the farmer determines that the sharp object is a needle. The analysis enables the farmer to find and remove the needle whenever it occurs in a haystack.
In a separate situation, a farmer is perplexed, having no idea why his haystacks are rotting so fast. The problem is causing him to lose a lot of revenue he makes from selling the hay. Examining vast quantities of data, the deep learning analysis connects the arrival of a transport truck from Al’s Milk Service to the presence of water rot in haystacks, which are covered when it rains. It also determined that the problem only occurs the day after heavy rains. Using the insight, the farmer notices that heavy rain creates a large and deep puddle next to his haystacks. He also observes that the day after it rains, the milk truck drives over the puddle, which hasn't yet evaporated, and splashes water into the haystacks, which are now uncovered as the rain has stopped. The water from the puddle is causing the rot. The farmer puts in drainage and the damaging problem never occurs again.
Okay, so I'm not a farmer, but you get the point. The former example showed known insight that was hidden. The latter method showed previously unknown insight that would never had been discovered. Both are valuable, as they can both identify a issues causing critical problems. However, they take different paths to accomplish objectives. Organizations looking to use either types of analytics need to first consider what problems they are trying to solve and how much they now about the underlying causes.
The second part of this blog series will take a closer look at four specific cognitive analytics capabilities that can be used to incorporate unstructured data, using the example of asset failure prediction as part of an overall asset performance management (APM) strategy.