Accuracy refers to the number of training examples that are correctly valued/classified as a proportion of the total number of examples in the training set. More...

#include <model_metric.h>

Inherits vita::model_metric.

Public Member Functions
double	operator() (const core_class_lambda_f *, const dataframe &) const override

double	operator() (const core_reg_lambda_f *, const dataframe &) const override

virtual double	operator() (const core_class_lambda_f *, const dataframe &) const =0

virtual double	operator() (const core_reg_lambda_f *, const dataframe &) const =0

Detailed Description

Accuracy refers to the number of training examples that are correctly valued/classified as a proportion of the total number of examples in the training set.

According to this design, the best accuracy is 1.0 (100%), meaning that all the training examples have been correctly recognized.

Note: Accuracy and fitness aren't the same thing. Accuracy can be used to measure fitness but it sometimes hasn't enough "granularity"; also it isn't appropriated for classification tasks with imbalanced learning data (where at least one class is under/over represented relative to others).

Definition at line 55 of file model_metric.h.

Member Function Documentation

◆ operator()() [1/2]

double vita::accuracy_metric::operator()	(	const core_class_lambda_f *	l,
		const dataframe &	d
	)		const

overridevirtual

Parameters

[in]	l	the model whose accuracy we are evaluating
[in]	d	a dataset

Returns: the accuracy

Implements vita::model_metric.

Definition at line 50 of file model_metric.cc.

◆ operator()() [2/2]

double vita::accuracy_metric::operator()	(	const core_reg_lambda_f *	l,
		const dataframe &	d
	)		const