how to manually adjust threshold value in PERFCURVE(​LABELS,SCO​RES,POSCLA​SS)

I have 100 subjects, I used KNN classifier for classification, when i use perfcurve function for roc it automatically displays result, the question in my mind is how and where to adjust threshold values in perfcurve that causes on result.

[Mohammad Sami]

There are addtitional returned values from the perfcurve function. For more details check the documentation.

The third returned values T, [X,Y,T] = perfcurve(labels,scores,posclass) returns an array of thresholds on classifier scores for the computed values of X and Y. You can then choose the optimal value of T from this array based on your needs.

[KALYAN ACHARJYA]

I am following this question

[Balaji M. Sontakke]

Sir, as you said "You can then choose the optimal value of T from this array based on your needs.", But in my perfcurve, I get 11 values start from 1,2,3......10.

many questions are comes in mind regarding this threshold, I will ask one by one here...

1) How the researchers get threshold like in following table but in my result only I get threshold like 1,2,3....10

Threshold FAR (%) FRR (%) EER (%) GAR (%)

0.1 15.61 42.06 28.84 71.15

0.13 20.30 22.66 21.48 78.51

0.2 31.23 8.73 19.98 80.01

0.3 46.85 2.93 24.89 75.10

0.4 62.46 2 32.23 67.76

0.5 78.08 1.4 39.74 60.25

Following is the result of my program...

Confusion matrix

4 0 0 0 0 0 0 0 0 0

0 3 0 0 0 0 0 1 0 0

1 1 0 0 0 1 0 0 1 0

0 1 1 2 0 0 0 0 0 0

0 0 0 1 1 0 0 0 1 1

1 0 0 0 0 0 1 2 0 0

2 0 0 0 0 0 2 0 0 0

1 0 0 0 0 1 0 2 0 0

1 0 0 1 0 0 0 0 2 0

0 1 2 0 0 0 0 0 0 1

Total Instance = 40

class1==>1

class2==>2

class3==>3

class4==>4

class5==>5

class6==>6

class7==>7

class8==>8

class9==>9

class10==>10

Multi-Class Confusion Matrix Output

TruePositive FalsePositive FalseNegative TrueNegative

____________ _____________ _____________ ____________

Actual_class1 4 6 0 30

Actual_class2 3 3 1 33

Actual_class3 0 3 4 33

Actual_class4 2 2 2 34

Actual_class5 1 0 3 36

Actual_class6 0 2 4 34

Actual_class7 2 1 2 35

Actual_class8 2 3 2 33

Actual_class9 2 2 2 34

Actual_class10 1 1 3 35

Overall values

Accuracy: 0.4250

Error: 0.5750

Sensitivity: 0.4250

Specificity: 0.9361

Precision: 0.4467

FalsePositiveRate: 0.0639

F1_score: NaN

MatthewsCorrelationCoefficient: 0.3900

Kappa: 0.6870

Tbl =

11×3 table

FPR TPR Thr

_______ ___ ___

0 0 10

0.13333 0 10

0.23333 0.1 9

0.33333 0.2 8

0.4 0.4 7

0.5 0.5 6

0.63333 0.5 5

0.76667 0.5 4

0.86667 0.6 3

1 0.6 2

1 1 1

[Mohammad Sami]

Use the roc function for the multiclass problem.

An example is shown in the documentation.

https://www.mathworks.com/help/releases/R2019b/deeplearning/ref/roc.html