The first submitted models to binary questions have got very low AUCs. However, if they try "1- predicted values" results will improve alot. For example AUC 0.25 can increase to 0.75! Is this procedure acceptable?

Created by S J SAJA
Thanks!
You may use any method for your predictions, including learning from your "test" submissions as long as you restrict yourself to predictors in the appropriate time ranges. Your LOOCVs must use the same method as your final predictions as must your permutations if you choose to submit them, and your methods must be documented.

Question about binary predictions page is loading…