How to identify overfitting or underfitting in ANN model?

I have simulated ANN model and also obtained the R values corresponding to training, testing, validation and all data. Now how can i scientifically identify that model is overfitted or underfitted?

Is there any scientific criteria for this.

Kindly help.

Answer 1

[Piyush Kumar]

@Rahul Verma,

Identifying overfitting and underfitting in an Artificial Neural Network (ANN) model involves analyzing the model’s performance on training, validation, and test datasets.

Indicators of Overfitting and Underfitting -

1. High Training Accuracy, Low Validation/Test Accuracy: If the model shows significantly higher accuracy on the training set compared to the validation or test set, it is likely overfitting.
2. Low Training and Validation/Test Accuracy: If the model performs poorly on both the training and validation/test sets, it is likely underfitting.

Methods to detect Overfitting and Underfitting -

• Learning Curves: Check this medium article that describes how to use "Learning Curve to identify Overfitting and Underfitting in a model".