If you think the subject of predictive analytics must be as dull as a math story problem, think again. The power of predictive analytics will blow your mind. Richard Holada, vice president of BI/AA for IBM’s Software Group, recently gave a group of fleet executives at the FleetRisk Advisors’ annual user group meeting an insider’s view of the world of big data and predictive analytics. Even for a math-averse wordsmith like myself, his presentation was riveting, like peeking under the curtain of the future.
“I work with customers around the world,” Holada said. “IBM has a global exposure concerning predictive analytics. ” Global clients include airlines, government agencies, telecomm giants, automakers, the medical community, and insurance providers, as well as truck fleets. And the problems predictive analytics is helping to solve are as various as the clients themselves. Holada cited several examples:
- Reducing infection-caused mortality rates after heart valve replacement by 6%
- Helping a car company predict when engine block molds will go out of tolerance due to wear so that they can be replaced before blocks come out bad
- Enabling an insurance company to identify the few customers who are trying to cheat the system so that they can provide better, faster service to everyone else—and reduce costs by not having to investigate every customer
- Identifying the string of events that lead up to airline catastrophes in order to better prevent accidents
- Pinging some 5,000 sensors onboard every car every hour for a luxury automaker to identify anomalies that may cause problems down the road
Predictive analytics is mathematics on steroids. FICO Analytic Solutions says, “Predictive analytics encompasses a variety of mathematical techniques that derive insight from data with one clear-cut goal: Find the best action for a given situation.” And that is about as clean a definition as one could hope to find.
Things really start to get interesting, however, when the amount of data pushes up into the millions and even billions of points—so much data that it takes parallel computing power to process it and algorithms I certainly never met in a math class to make sense of it all.
Add predictive modeling to the picture and you are, amazingly, peeking under the curtain of the future. So what is predictive modeling? FleetRisk Advisors notes that it is “a part of predictive analytics and is used to determine the probable future outcome of an event or the likelihood of a situation occurring by analyzing large amounts of historical data. By using pattern recognition, software modelers are able to identify the recurring patterns in past events and then use those patterns to develop a model to predict the near future.”
FleetRisk Advisors is using predictive modeling to help identify which drivers are most apt to have an accident in the short term in time to intervene. Others are using it to identify components that are ready to fail in time to replace or repair parts before a breakdown occurs. The opportunities are as endless as the list of problems truck fleets can encounter—and that is a long list indeed.
Holada offers his own matter-of-fact view of this almost magical process: “Predictive analytics is all about taking a business problem [and addressing it],” he said. “We pick a point in a process flow where, if we could be predictive, it would make a lot of difference.” No kidding.
