R2 / Coefficient of Determination

Theory

Used to determine strength of correlation between features and target.

SSR (unexplained variation)

SST (total variation)

R2 of 1, means 100% of the variation is explained.

R2 will be 0, if the SSR=SST; which means the model is not performing better than the average.

Adjusted R squared

R2 will increase by adding more numbr of features; r2 inflation. As R2 always increases, you might always add more features; which is misleading.