From the course: Problem Identification and Solution Design for Data Scientists

Unlock the full course today

Join today to access over 24,900 courses taught by industry experts.

Diving deeper into predictive analytics

Diving deeper into predictive analytics

From the course: Problem Identification and Solution Design for Data Scientists

Start my 1-month free trial Buy for my team

Diving deeper into predictive analytics

“

- [Instructor] Okay now, something kind of subtle, but very important. You need to ask them when they need to know what they're trying to predict. At first, you might think the answer is, as soon as possible, and that's likely the first thing they'll say, but then you have to explain more carefully. Let me give you an example. Let's say you want to build a model to fast track insurance claims. If there's a very low risk of fraud, maybe a tree limb came down in a storm, they haven't had a claim in 17 years, and they have pictures, versus a roof damage claim for an older roof. The roof claim may be valid, but you have to send an inspector out. So for the fast track model, it should be built to make a decision within a short period of time after the claim, conceivably in real time while they're still on the phone. And that also means that the model is restricted to the information that is known at that time, whereas the fraud model for the roof has to wait for the inspection, and it…

Contents