I just finished training a CNN model for binary classification. Two diagrams were created afterward, but as I am very new to machine learning, I don't get what they state. can anyone tell me if my model has good performance.
Thank you.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
These two charts are showing how the accuracy and loss changed with each epoch.
This model is trained for 10 epochs and the first chart is showing the training accuracy and validation accuracy from the start of the training. Usually, the training accuracy increases as epochs increase.
The validation accuracy also should increase but with more epochs, it might decrease indicating the best model. This decrease in validation accuracy after a certain point indicates "Overfitting" and you should stop training there.
The second chart is keeping track of the loss. You might have defined binary_crossentropy or categorical_crossentropy loss depending on the number of classes for classification. So, ideally, your loss should keep decreasing as epochs increase.
Currently, your model has a validation accuracy of 83-84%.
And telling whether the model is working good depends on what accuracy is expected. In some cases, 70% is considered to be good enough and in some cases, 90% is also not considered as good. It solely depends on the nature of the problem you are solving.
You can try more no.of epochs, and try to add more data if it is feasible.
Also, add information regarding what is the training data size, train test split size, model details (summary), no.of classes, batch size etc.
You can also play around with these parameters or what is known as "Hyper-parameter tuning" to obtain better results.
