search

aflondono / mlf_capstone

Capstone project for the Codecademy Machine Learning Fundamental Intensive course
0 stars 0 forks source link

R-squared is not a good measure of accuracy when using regression for binary classification #2

Open alexander-dubinski opened 5 years ago

alexander-dubinski commented 5 years ago

When you are trying to predict something that is binary (I.E. gender) the r-squared will usually be low and is not a good measure of suitability. As with other classification techniques, a good measure of accuracy in this case is to have a hold-out subset and test to see if they are predicted accurately. The percent which was correctly predicted is the best measure of accuracy when using regression for two class classification.

aflondono commented 5 years ago

@addubinski This is a good point, and obviously I didn't think of that. Thanks.