Predictive - Logistic Regression Error

[Su777]

Good evening everyone!

Can anyone provide feedback to the following error of the Logistic Regression tool. The images are shown below:

Logistic Regression ToolTarget variable - Creditworthy Applicant

[Inactive User · Accepted solution]

9/10 times it’s because you have (1) missing values / nulls, (2) numeric values formatted as strings or (3) variables without any variation (e.g. all 1’s)

I see that your “account-balance” is a string. Sure it’s not a numeric variable?
Also, Occupation is numeric. Sure it’s not a string?

[Charity_K_Wilson · Accepted solution]

Most of your variables are String.  If the independent variables are binary, try switching them over to double. 

A logit regression works much better when the Independent variables are actually integers.  The model is more robust when things are continuous rather than categorical. 

[Su777]

Occupation is a string but it needed to be excluded as a predictor variable. It was originally included in the analysis for the Logistic Regression and so was the account balance.

After some help from Matt David at Udacity and my peers in the my class, I was able to correctly select the correct variables and it ran without any problems! Yay!

But thank you for your tip and I'll be sure to keep it in mind. :D

[Su777]

Hi Charity_K_Wilson,

A few days after I posted this thread, I figured out what was wrong with the LR and, yes, it had something to do with the variables. Once I excluded certain ones, it ran without any problems!

Thank you for the tip, however, and I will certainly remember it next time I use the LR.

Su