From the course: Data Science Foundations: Data Assessment for Predictive Modeling

Unlock the full course today

Join today to access over 22,400 courses taught by industry experts or purchase this course individually.

Advice for weak predictors

Advice for weak predictors

From the course: Data Science Foundations: Data Assessment for Predictive Modeling

Start my 1-month free trial

Advice for weak predictors

- [Instructor] Okay, we're going to continue with the same dataset and the same workflow but turn our attention to weak predictors. So I'll close this. So when we're looking at predictors, we're going to have to do more detective work. The way most people behave is they treat the data understanding phase as a kind of screening phase and they get rid of variables left and right. We don't want to do that. I want to encourage a different approach. What you want to do is when there's a variable that has problems, dig deep and see if you can't save it. See if there are interesting relationships there. So for instance, the variable that comes to mind for having problems would be variables like capital-gain and capital-loss. We'll look at capital-gain. The reason is there's so many zeros. Let's run this and talk about it. That large number of zeros isn't necessarily going to bring the variants and standard deviation way…

Contents