Validation

Full name Microsoft.VisualBasic.DataMining.Evaluation.Validation Assembly Microsoft.VisualBasic.DataMining.Framework Members 28

验证结果描述

灵敏度 = 真阳性人数 / (真阳性人数 + 假阴性人数) * 100% 特异度 = 真阴性人数 / (真阴性人数 + 假阳性人数) * 100%

00 Remarks

https://www.jianshu.com/p/f0c7c1ad9091

01 Syntax

Microsoft.VisualBasic.DataMining.Evaluation.Validation

02 Methods

Name	Overloads	Summary
Calc	1
ROC	1	生ROC曲线的绘制数据(这个函数产生的曲线默认是阈值在[0,1]之间的)
ToString	1
ToDataSet	1
Calc	1
AUC	1
ROC	1

03 Properties

Name	Overloads	Summary
NPV	1	Negative predictive value
FPR	1
F1Score	1
FbetaScore	1

04 Fields

Name	Overloads	Summary
Specificity	2	TNR
Sensibility	2	Recall, TPR
Precision	2	PPV
BER	2	balanced error rate
Threshold	2	进行当前的预测鉴定分析的百分比等级，默认是0.5，即 50%
Accuracy	1
All	1
TP	1
FP	1
TN	1
FN	1
normalRange	1

05 Members

method Calc #

Calc``1(IEnumerable(Of ``0), Func(Of ``0, Boolean), Func(Of ``0, Boolean), Double)

Type Parameters

Name	Description
`T`	`true` 表示阳性 `false` 表示阴性

Parameters

Name	Type	Description
`entity`	IEnumerable(Of ``0)	-
`getValidate`	Func(Of ``0, Boolean)	得到实际的分类结果
`getPredict`	Func(Of ``0, Boolean)	得到预测的分类结果

method ROC #

ROC``1(IEnumerable(Of ``0), Func(Of ``0, Double, Boolean), Func(Of ``0, Double, Boolean), Variant(Of Sequence, Func(Of ``0, Double)))

生ROC曲线的绘制数据(这个函数产生的曲线默认是阈值在[0,1]之间的)

Remarks

在一个二分类模型中，对于所得到的连续结果，假设已确定一个阈值，比如说 0.6，大于这个值的实例划归为正类，小于这个值则划到负类中。如果减小阈值，减到0.5，固然能识别出更多的正类，也就是提高了识别出的正例占所有正例的比例，即TPR，但同时也将更多的负实例当作了正实例，即提高了FPR。为了形象化这一变化，在此引入ROC。

Type Parameters

Name	Description
`T`

Parameters

Name	Type	Description
`entity`	IEnumerable(Of ``0)	-
`getValidate`	Func(Of ``0, Double, Boolean)	`func x, threshold => yes/no`
`getPredict`	Func(Of ``0, Double, Boolean)	-

property NPV #

NPV

Negative predictive value

field Specificity #

Specificity

TNR

field Sensibility #

Sensibility

Recall, TPR

field Precision #

Precision

PPV

field BER #

BER

balanced error rate

field Threshold #

Threshold

进行当前的预测鉴定分析的百分比等级，默认是0.5，即 50%

field Specificity overload 2 #

Specificity

field Sensibility overload 2 #

Sensibility

field Accuracy #

Accuracy

field Precision overload 2 #

Precision

field BER overload 2 #

BER

field All #

All

field TP #

field FP #

field TN #

field FN #

field Threshold overload 2 #

Threshold

field normalRange #

normalRange

property FPR #

FPR

property F1Score #

F1Score

property FbetaScore #

FbetaScore

method ToString #

ToString()

method ToDataSet #

ToDataSet()

method Calc #

Calc(IEnumerable(Of T), Func(Of T, Boolean), Func(Of T, Boolean), Double)

method AUC #

AUC(IEnumerable(Of Validation))

method ROC #

ROC(IEnumerable(Of T), Func(Of T, Double, Boolean), Func(Of T, Double, Boolean), Variant(Of Sequence, Func(Of T, Double)))