Predictive Pitfalls

The Dark Side of Predictive Analytics

As a scientist, you're probably no stranger to making predictions about the world around you.

But beware: predictions can be a slippery slope, and the pitfall-ridden path that lies ahead is full of surprises.

Here are just a few of the many pitfalls that come with the territory:

1. The Overfitting Overlord

A model that's too good at its job may be a model that's too good at fitting the noise, rather than the signal.

Read more

2. The Selection Bias Snake Pit

When you're only looking at a subset of data, you're only getting part of the story – and that's a recipe for disaster.

Slither into the details

3. The Data Snooping Scandal

You can't make predictions if you don't have the data – but if you do have the data, be prepared for snooping and snooping.

Snoop for more
Back to Scientists