SETCUROP Set current operating point in the ROC object [R,IND]=SETCUROP(R,IND) [R,IND]=SETCUROP(R,'min',MEASURE) [R,IND]=SETCUROP(R,'max',MEASURE) Cost-sensitive optimization [R,IND]=SETCUROP(R,'cost',M) Combine constraining and setting op.point (see full example below) [R,IND]=setcurop(R,'constrain','TPr(banana)',0.8,'min','FPr(banana)') INPUT R SDROC object IND Index of operating point to be set MEASURE Measure specification (index or name i.e. 'err(apple)') M Cost matrix (R must store confusion matrices i.e. be created with 'confmat' option) OUTPUT R SDROC object with the cur.op.point set IND Index of the selected operating point DESCRIPTION SETCUROP sets the current operating point in the SDROC object R. The operating point may be specifed by its index IND or based on the related performance measure value. Any measure may be maximized or minimized. SETCUROP may also perform cost-sensitive optimization. When provided with the cost matrix M, SETCUROP selects operating points minimizing classifier loss function. To perform cost-sensitive optimization, R must contain full confusion matrices ('confmat' option). SETCUROP allows to directly apply constrain and set an operatin point in one command. In order to do that, use 'constrain' as the first option with the name of a measure and numerical value (see syntax of contrain method for details). After, you need to use 'min','max' or 'cost' option to fix operating point. EXAMPLE >> load fruit; b=a(:,[2 3]) >> p=sdfisher(b) >> r=sdroc(b,p,'measures',{'FPr','banana','TPr','banana'} >> r=setcurop(r,'constrain','TPr(banana)',0.95','min','FPr(banana)') The last command applies constrain and sets op.point directly in full object r SEE ALSO SDROC, SDLOSS, GETCUROP, CONSTRAIN