Use visual plot of hand-made decision rules rather than decision tree plot

Module 1. The Predictive Modeling Pipeline Tabular data exploration
#1

Here are a few remarks about this module :

  1. When you plot the data using seaborn, you discuss hand-written rules which do not seam that obvious to me (for instance " if you are old (more than 70 year-old roughly), you are in the <=50K class). Maybe it would be interesting to point out in which of the 9 plots you get the information you’re talking about

  2. Regarding the Example of machine learning model decision rules :

  • the bottom axis of the figure is cropped
  • maybe it would be better to make us plot this figure instead of showing a screenshot
  • When going through the module, I was kind of confused with this last block. It felt out of place, as the module is supposed to take us through the basics of data exploration and visualization. There is too much new information, with no background knowledge and no interactivity.
#2

the bottom axis of the figure is cropped

This was noted elsewhere (Konrad Hinsen IIRC) we should definitely fix this.

maybe it would be better to make us plot this figure instead of showing a screenshot

the problem is that this is using a scikit-learn model and we don’t want to explain this code that would be quite intimidating so probably good enough as it is.

When going through the module, I was kind of confused with this last block. It felt out of place, as the module is supposed to take us through the basics of data exploration and visualization. There is too much new information, with no background knowledge and no interactivity.

OK I find it a bit :cry: but honest feed-back is really appreciated, that is what the beta is for. You are right that this is kind of not in the main notebook topic. The idea is to give some insights that a machine learning can create rules that would be similar to what a human would see without having to write scikit-learn code.

If you have some suggestions about what could improve the situation, that would be more than welcome :pray:

#3

I don’t know…I think just getting rid of the last block would be just fine.
There will be plenty of interactive examples of how to make predictions later in the mooc.

Say a few words about decision tree models the first time they are introduced
#4

I think we decided to not talk about a decision tree model but rather have a visual plot for what hand-made rules would look like.

General feedback/various points
#5

Done in github.

#6