Logistic Regression : quantitative / qualitative predictor variables and field type

[simonaubert_bd]

Hello all,

I'm working on a very simple logistic regression workflow but I'm a beginner here and I must say the help isn't that simple.

Data is like this

Order_Id	Country	Country_id	Category	Category_id	Amount	Discount	At_Risk
1	Germany	1	Bike	1	1100	0	1
2	France	2	Bike	1	2200	0	0
3	Germany	1	Clothes	2	30	0	0
4	Spain	3	Bike	1	1290	0,2	0
5	France	2	Bike	1	1750	0	0
6	Germany	1	Bike	1	2100	0,3	1
7	Spain	3	Accessories	3	125	0	0
8	Portugal	4	Bike	1	1800	0	0

My target variable is At_Risk.

The first thing I have noticed is that you can't use a numeric variable as a target variable, I don't know why and how that makes sense. Only string variables are proposed in the dropdown menu.

But ok.

Then, I have a few categorical variable (like Country or Category) and also something more quantitative (amount). It seems that I don't have to distinguish it? Also, I can use both strings and numbers in that ? The only thing I have to avoid is almost duplicate variable like Country and Country_id.

Am I right?

is there something I miss ? I wonder if I should make categories of amount instead of having the exact amount.

Thanks for your help.

Simon

[aatalai]

@simonaubert_bd to respond to your first statement, re-target variable being only a string. Logistic regression is for classification (and how it is set up in most cases and in alteryx can only have 2 outcomes, a yes/no, true/false,0/1) hence why the target variable needs to be a string rather than numerical. 

You wouldn't want variables that represent the same information for example country and country id as you suggested.

If you want an exact amount can I suggest  numerical model, e.g. Liner regression, Random forrest, DT.