Hi Omar 

the attached matlab function  has as input

- pred_val: prediction values (0,1)

- yval: actual values (0,1) 

- verbose: verbose mode (default = 1)

and it returns typical metrics of classifiers (among whom accuracy):

 - F1: F1 score 

- precision: precision  

- recall: recall 

- accuracy: accuracy 

HTH

thank you for your helpful answer

You can use recall and accuracy, the use them to calculate the F-Measure. You can use the primitive like true positive, true negatives, false positives, false negatives if your classification is binary.

Matlab links:

ROC curves:

http://es.mathworks.com/help/stats/perfcurve.html;jsessionid=8dc31085c8727b39e439edb20577?refresh=true

thanks alot for you

The simplest way is the cross validation method using the training and test sets.

Sir 

I want to know that while applying confusion matrix if matrices are not of same dimensions then what can be done.

Hi;

This is code in MATLAB to calculate all FACTORS .

Best Regards

Seyed Yahya Moradi Thank you for your reply. But i got error when i run the code :

No constructor 'handle.listener' with

matching signature found.

Error in rotateXLabels/addListeners (line

321)

listeners = [

Error in rotateXLabels/repositionAxes

(line 220)

addListeners( ax );

Error in rotateXLabels (line 41)

repositionAxes( ax );

Error in draw_cm (line 27)

rotateXLabels(gca, 315 );% rotate the x

tick

Error in compute_confusion_matrix (line

23)

draw_cm(confusion_matrix,name_class,num_class);

Error in main (line 19)

[confusion_matrix]=compute_confusion_matrix(predict_label,num_in_class,name_class);

Check this link for your query.

https://www.mathworks.com/help/deeplearning/examples/classify-sequence-data-using-lstm-networks.html

There are many parameters for validation of a calssifier.

Accuracy is one of these parameters.

So what it means? Accuracy determines that how percentage of test data is correctly classified.

It can be calculated accurding to this equation :

Accuracy= ( number of true classified samples)/ ( number of total test data) × 100;

So how to calculate this in matlab?

Suppose d is our test data's true labels;

And p is your models predicted labels for test data,

Then we can simply calculate accuracy as follows :

Accuracy = sum ( d==p) / numel (d) ×100

We can also calculate accuracy from confusion matrix as follows :

If C is your confusion matrix, then:

Accuracy = sum ( diag (C)) / sum (C (:)) ×100

First, you should have true positives, false positives, and false negatives then you can calculate Accuracy or F1-score, check these for formulas

https://en.wikipedia.org/wiki/F-score

MCC

Accuracy

F Score