meta::classify::confusion_matrix Class Reference

Allows interpretation of classification errors. More...

#include <confusion_matrix.h>

Public Types
typedef std::unordered_map< std::pair< class_label, class_label >, size_t, decltype(&confusion_matrix::string_pair_hash)>	prediction_counts
	typedef for predicted class assignments to counts. More...

Public Member Functions
	confusion_matrix ()
	Creates an empty confusion matrix.
void	add (const class_label &predicted, const class_label &actual, size_t times=1)
void	print_stats (std::ostream &out=std::cout) const
	Prints this matrix's statistics to out. More...
void	print (std::ostream &out=std::cout) const
	Prints this matrix to out. More...
void	print_result_pairs (std::ostream &out=std::cout) const
	Prints (predicted, actual) pairs for all judgements. More...
const prediction_counts &	predictions () const
double	accuracy () const
confusion_matrix	operator+ (const confusion_matrix &other) const
	operator+ for confusion matrices. More...
confusion_matrix &	operator+= (const confusion_matrix &other)
	operator+= for confusion matrices. More...

Static Public Member Functions
static size_t	string_pair_hash (const std::pair< std::string, std::string > &strPair)
	Implements a hash function for a pair of strings. More...
static bool	mcnemar_significant (const confusion_matrix &a, const confusion_matrix &b)

Private Member Functions
void	print_class_stats (std::ostream &out, const class_label &label, double &prec, double &rec, double &f1, size_t width) const
	Prints precision, recall, and F1 for each class and as a whole. More...

Private Attributes
prediction_counts	predictions_
	maps predicted class to actual class frequencies
std::set< class_label >	classes_
	Keeps track of the number of classes. More...
std::unordered_map< class_label, size_t >	counts_
	how many times each class was predicted
size_t	total_
	total number of classification attempts

Detailed Description

Allows interpretation of classification errors.

Member Typedef Documentation

typedef std::unordered_map<std::pair<class_label, class_label>, size_t, decltype(&confusion_matrix::string_pair_hash)> meta::classify::confusion_matrix::prediction_counts

typedef for predicted class assignments to counts.

Member Function Documentation

void meta::classify::confusion_matrix::add	(	const class_label &	predicted,
		const class_label &	actual,
		size_t	times = 1
	)

Parameters

predicted	The predicted class label
actual	The actual class label
times	The number of times this prediction was made

void meta::classify::confusion_matrix::print_stats

(

std::ostream &

out = std::cout

)

const

Prints this matrix's statistics to out.

Parameters

out	The stream to write to (defaults to std::cout)

void meta::classify::confusion_matrix::print

(

std::ostream &

out = std::cout

)

const

Prints this matrix to out.

Parameters

out	The stream to write to (defaults to std::cout)

void meta::classify::confusion_matrix::print_result_pairs

(

std::ostream &

out = std::cout

)

const

Prints (predicted, actual) pairs for all judgements.

Parameters

out	The stream to write to (defaults to std::cout)

size_t meta::classify::confusion_matrix::string_pair_hash ( const std::pair< std::string, std::string > &  strPair )

static

Implements a hash function for a pair of strings.

Parameters

str_pair

The pair of strings

Returns

the hash

const confusion_matrix::prediction_counts & meta::classify::confusion_matrix::predictions

(

)

const

Returns

all the predictions from this confusion_matrix.

double meta::classify::confusion_matrix::accuracy

(

)

const

Returns

the accuracy for this confusion matrix

confusion_matrix meta::classify::confusion_matrix::operator+

(

const confusion_matrix &

other

)

const

operator+ for confusion matrices.

All counts are agglomerated for all predictions.

Parameters

other

Returns

a confusion_matrix containing all predictions of the parameters

confusion_matrix & meta::classify::confusion_matrix::operator+=

(

const confusion_matrix &

other

)

operator+= for confusion matrices.

All counts are agglomerated for all predictions.

Parameters

other

Returns

a confusion_matrix containing all predictions of the parameters

bool meta::classify::confusion_matrix::mcnemar_significant ( const confusion_matrix &  a, const confusion_matrix &  b  )

static

Parameters

a	The first matrix to compare
b	The second matrix to compare

Returns

whether results between two confusion matrices are statistically significant according to McNemar's test with Yates' correction for continuity (alpha = .05)

void meta::classify::confusion_matrix::print_class_stats ( std::ostream &  out, const class_label &  label, double &  prec, double &  rec, double &  f1, size_t  width  ) const

private

Prints precision, recall, and F1 for each class and as a whole.

Parameters

out	The stream to print to
label	The current class label to get statistics for
prec	The precision for this class
rec	The recall for this class
f1	The F1 score for this class

Member Data Documentation

std::set<class_label> meta::classify::confusion_matrix::classes_

private

Keeps track of the number of classes.

We use a std::set here so the class labels are sorted alphabetically.

---

The documentation for this class was generated from the following files:

• /home/chase/projects/meta/include/classify/confusion_matrix.h
• /home/chase/projects/meta/src/classify/confusion_matrix.cpp