[SOLVED] Optimizing Metacost Matrix operator

jforr

I'm trying to use the metacost operator to overguess on one of my labels, while keeping the false positives to a reasonable level.  Is there a way of optimizing the matrix inputs?

Alternatively, are there any other general suggestions for optimizing recall at the expense of precision for a particular label in a binary label dataset?

MariusHelf

Hi,

if you know which recall you want to achieve, you can use Select Recall and Apply Threshold.

You can't set the values of the cost matrix directly; however you could substitute them by macros, place Set Macro operators before the MetaCost operator (inside the Optimization operator), and optimize the value fields of the Set Macro operators.

Best,
Marius