GEP::ClassificationIndividual Class Reference

Individual performing classification tasks It is used for binary classification (having two classes). It holds four Predicted/Actual values. During evaluation, chromosome value is compared with the target, and classification matrix (below) is built, used for fitness calculation. More...

#include <Individual.h>

Inheritance diagram for GEP::ClassificationIndividual:

[legend]

Collaboration diagram for GEP::ClassificationIndividual:

[legend]

List of all members.

Public Member Functions
	ClassificationIndividual ()
	ClassificationIndividual (const ClassificationIndividual &)
virtual	~ClassificationIndividual ()
bool	classify ()
void	evaluate (const u_short &user_param=0.0, const size_t pop_age=0)
ret_type	calculateFitness (const size_t pop_age=0)
void	reset ()
std::string	toString () const
Public Attributes
double	TP
double	FN
double	FP
double	TN
bool	zero_output
Static Public Attributes
static double	roundingThreshold = 0.5
static double	weightFP = 1
static double	weightFN = 1

Detailed Description

Individual performing classification tasks It is used for binary classification (having two classes). It holds four Predicted/Actual values. During evaluation, chromosome value is compared with the target, and classification matrix (below) is built, used for fitness calculation.

     Predicted
     A      |  Yes |  No
     c   ---+------+-------	
     t   Yes|  TP  |  FN	 
     u   ---+------+-------
     a   No |  FP  |  TN
     l

Constructor & Destructor Documentation

GEP::ClassificationIndividual::ClassificationIndividual ( )

Default constructor

GEP::ClassificationIndividual::ClassificationIndividual ( const ClassificationIndividual & copy )

Copy constructor

virtual GEP::ClassificationIndividual::~ClassificationIndividual ( ) [inline, virtual]

Default destructor

Member Function Documentation

double GEP::ClassificationIndividual::calculateFitness ( const size_t pop_age = 0 ) [virtual]

Uses specificity*sensitivity measure:

 _fitness = ((TP/(TP+FN))*(TN/(TN+FP))) * 1000.

Maximum fitness = 1000.

Reimplemented from GEP::Individual< double, double, u_short >.

bool GEP::ClassificationIndividual::classify ( )

Classification uses rounding threshold to obtain predicted result (0 or 1) (see evaluate()).

void GEP::ClassificationIndividual::evaluate	(	const u_short &	user_param = `0.0`,
		const size_t	pop_age = `0`
	)			`[virtual]`

Evaluation uses classify() output which is then compared with actual output passed in user_param (actual).

   if (_chromosome.getValue() < roundingThreshold) predicted = 0; else predicted 1;
   if(actual == 0) { //Positive
   if(predicted == 1) FP++;
   else TP++;
   }
   else { //Negative
   if(predicted == 1) TN++;
   else FN++;
   }