I heard about a fascinating use of predictive analytics recently. The company concerned is DriveCam, who focus on improving risky driving behavior by predicting and preventing crashes. They provide an exception-based video event recorder that captures sights and sounds inside and outside a vehicle. G-forces (e.g. hard braking, swerving, collision, etc.) cause the recorder to save the 10 seconds immediately before and after the triggered event. This clip is routed to the operation center for review were a risk analyst scores the clip for risky behavior. If the clip scores high enough then a "coach" reaches out to a fleet manager (or the parent of a teenage driver) and makes recommendations to change the driver’s behavior. The idea is to use video clips of a driver to modify their behavior BEFORE they get into an accident.
Now accurately pinpointing and weighting contributing factors is a daunting task and DriveCam is using predictive analytics to sift through the mass of data to more effectively and accurately identify patterns of behavior. In particular, obviously, to predict which patterns represent higher risk driving. These predictive analytics can then be used to identify trends in the data from each clip to help find the ones that should be reviewed and acted on. Not only does this help DriveCam do its job better, it also puts in place a "decision service" that can be constantly improved. Experience is that the analytics will likely become more effective as more is learned, reducing the rate at which new staff must be added to handle business growth.
You can hear more about DriveCam in their presentation"Cutting-Edge Analytics using Audio and G-Force Data" as part of the analytics track at InterACT. Don't forget there is a blog discount code for InterACT (CRJT90)
P.S. You could imagine adding this kind of technology to improve Pay-As-You-Drive Insurance.