Most people use logistic regression for modeling response, attrition, risk, etc. And in the world of business, these are usually rare occurences.   One practise widely accepted is oversampling or undersampling to model these rare events. Sometime back, I was working on a campaign response model using logistic regression. After getting frustrated with the model performance/accuracy, I use weights…