Wrap-up Quiz 4

Module 4. Linear Models Wrap-up quiz
#1

I would like to know how should we identify the important feature from the boxplot.

Refer to the explanation done in this post,
https://lms.fun-mooc.fr/courses/course-v1:inria+41026+session01/courseware/f59f9342fd22439fb021fd580afd962c/cf4aec8730d1463093719f98c8698306/

If the box for a coefficient is large, it means it varies a lot with the randomization of the data. So for instance, with a coefficient that would vary between 0 and 100, 0 would mean that the feature will not be considered to predict the target while with 100, it will have an influence (to be more precise regarding the impact of this feature, we would also need to look at the other coefficient values).

My understanding is that large box plot means that the feature is able to have predictive power to the model.

In this case, we should choose the feature with a large boxplot. Thus, GarageCars and GarageArea meet the criteria.

Hope to get some explanation.

#2

No, a large box means that sometimes the weight is high, sometimes it is low. The box is just a way to visualize the weight variation. An important feature is linked if a large (positive or negative) weight.