Newbie: Logistic Regression Results interpretation

[thatsrajan]

Hi all,

Firstly my most humble apologies if this sounds stupid. I am new to predictive analytics and I am trying to get my head around the outputs from Alteryx. 

So I am using logistic regression at work, to determine the outcome of an "Alert" being escalated or not Escalated. I have picked the fields for Target and Predictor.

Attached are the results from the regression. What I do not understand is, how do I interpret the output from Port "I". Do these numbers in Recall mean that I have to set the "threshold to be 0.6" to get Optimal Recall ?

Also I know that the model is somewhat rubbish in-terms of fit as the R^2 is 0.05 and there seems to be no statistical significant for any of the coefficient. 

Am I heading in the right direction with regards to interpreting the data? 

Again apologies if I sound clueless. Thank you.

[fmvizcaino · Accepted solution]

Hi thatsrajan,

I'm far from an expert and not fluent in english, so be aware lol, but I'm going to try to answer your questions basically while no experts read this topic.

From the output I, the indicators are mainly calculations about your confusion matrix as image attached below.

The threshold parameter helps you into building your confusion matrix, so if you select a 0.6 threshold, every data above that score will be a predicted positive result and every data below will be a predicted negative result, you must select the best threshold depending on the indicator you want to follow as explained below (or at least tried to explain lol)

It depends on your business rules which indicator to mainly consider. Accuracy is a base indicator but depending on what you want, you should look into Recall or Specificity, the first if your concern is how well your predicting positives and the second if your concern is predicting negatives.

About the R^2, it is absolutely correct!

Best,

Fernando Vizcaino

[thatsrajan]

Hi fmvizcaino

Thank you so very much for this. It is so much clearer now.