ROCCurve Class Reference

ROCCurves show the trade-off in sensitivity and specificity for binary classifiers using different cutoff values. More...

#include <OpenMS/ML/ROCCURVE/ROCCurve.h>

Collaboration diagram for ROCCurve:

Classes
class	simsortdec
	predicate for sort() More...

Public Member Functions
	ROCCurve ()
	default constructor More...
	ROCCurve (const std::vector< std::pair< double, bool >> &pairs)
	constructor with value, class pairs More...
virtual	~ROCCurve ()=default
	destructor More...
	ROCCurve (const ROCCurve &source)
	copy constructor More...
ROCCurve &	operator= (const ROCCurve &source)
	assignment operator More...
void	insertPair (double score, bool clas)
	insert score, type pair More...
double	AUC ()
	returns Area Under Curve More...
double	rocN (Size N)
	returns ROC-N score (e.g. ROC-50). Returns -1 if not enough false positives were found More...
std::vector< std::pair< double, double > >	curve (UInt resolution=10)
	some points in the ROC Curve More...
double	cutoffPos (double fraction=0.95)
double	cutoffNeg (double fraction=0.95)

Private Member Functions
void	sort ()
	sorts data and caches if sorted More...
void	count ()
	counts global pos and neg More...
double	trapezoidal_area (double x1, double x2, double y1, double y2)

Private Attributes
std::vector< std::pair< double, bool > >	score_clas_pairs_
UInt	pos_ {}
UInt	neg_ {}
bool	sorted_ {}

Detailed Description

ROCCurves show the trade-off in sensitivity and specificity for binary classifiers using different cutoff values.

[This class is buggy and usage is discouraged!]

Constructor & Destructor Documentation

ROCCurve() [1/3]

ROCCurve

(

)

default constructor

ROCCurve() [2/3]

ROCCurve ( const std::vector< std::pair< double, bool >> &  pairs )

explicit

constructor with value, class pairs

~ROCCurve()

virtual ~ROCCurve ( )

virtualdefault

destructor

ROCCurve() [3/3]

ROCCurve

(

const ROCCurve &

source

)

copy constructor

Member Function Documentation

AUC()

double AUC

(

)

returns Area Under Curve

count()

void count ( )

inlineprivate

counts global pos and neg

curve()

std::vector<std::pair<double, double> > curve

(

UInt

resolution = 10

)

some points in the ROC Curve

cutoffNeg()

double cutoffNeg

(

double

fraction = 0.95

)

returns the score at which you would need to set a cutoff to have fraction positives returns -1 if there are not enough positives

cutoffPos()

double cutoffPos

(

double

fraction = 0.95

)

returns the score at which you would need to set a cutoff to have fraction positives returns -1 if there are not enough positives

insertPair()

void insertPair	(	double	score,
		bool	clas
	)

insert score, type pair

operator=()

ROCCurve& operator=

(

const ROCCurve &

source

)

assignment operator

rocN()

double rocN

(

Size

N

)

returns ROC-N score (e.g. ROC-50). Returns -1 if not enough false positives were found

sort()

void sort ( )

inlineprivate

sorts data and caches if sorted

trapezoidal_area()

double trapezoidal_area ( double  x1, double  x2, double  y1, double  y2  )

inlineprivate

calculates area with trapezoidal rule

Parameters

x1,x2,y1,y2

Member Data Documentation

neg_

UInt neg_ {}

private

pos_

UInt pos_ {}

private

score_clas_pairs_

std::vector<std::pair<double, bool> > score_clas_pairs_

private

sorted_

bool sorted_ {}

private