1) For each epoch, you should report the mean accuracy over the training samples and also the mean accuracy over the validation set. The comparison of both plots gives you an idea of the bias-variance trade-off.

2) The best way is to report the mean metrics (accuracy, recall, precision, F1). It's better to use cross-validation (e.g., 10x1, 5x2) so that you can perform statistical significance tests later or show box plots to compare different classifiers.

Hi Kannan u.m

There are several ways. I suggest calculating the confusion matrix with its kappa value.

Best regards,

Ebrahim

Thanks Alessandro Ferrarini , Giorgio Morales and Ebrahim Ghaderpour for your replies. I am already using the confusion matrix. Giorgio Morales can you please refer any publication for the F1 and cross-validation measurements. I would like to know more about those metrics

Kannan u.m , here's a reference on why it's a good idea to use cross-validation and how to perform statistical tests based on it: https://machinelearningmastery.com/statistical-significance-tests-for-comparing-machine-learning-algorithms/ . Cross-validation is not a metric but a method to evaluate a model.

For the F1 score, check section 5: https://www.cs.montana.edu/sheppard/pubs/jars-2020.pdf It's not a good idea to use only one metric but a combination of them so that you can explain if you're having more false positives or false negatives.

After the training phase is over and the weights have been generated, when a query item is fed at the input, a high value should be observed on one of the output lines of the network. This indicates the probability of the query data for belonging to one of the classes, and is the estimated classification. Comparing with the ground truth will generate the accuracy for a number of test items.

Thanks Giorgio Morales for the references. I will go through them

Thanks Ranjan Parekh for the answer

Dear, Kannan u.m!

I can develop a professional ANN (FANN) model for your task in python using KARAS and TensorFlow. Also, I can trained you this is skills. My contacts: [email protected] or Whatsapp(+79322301628).