Logistic Regression Classifier

User 3274 | 3/2/2016, 11:00:18 AM

I am struggling to determine the significant variables from the created logistic model. I have used 33 feature columns, I am not sure which ones out of those 33 are significant because there is no Wald's Chisq to suggest so. Kindly help.

Comments

User 1592 | 3/3/2016, 6:00:17 AM

Hi You can run model.get('coefficients').topk('value') to get the coefficient with the highest positive effect and model.get('coefficients').topk('value', reverse=True) to get coefficient with the highest negative effect.

Note that when using categorical or integer variables while starting from 33 features you can get a much higher number of features. I suggest viewing the following webinar: http://cc.readytalk.com/play?id=35zmsa that explains in more detail.