How to interpret the output of model.evaluate function?

[Nagabhushan Baddi]

Dear All,

   I have a fit a model of a binary classifier. The test set has 10002 examples. Here is the output of evaluate method:

>>> model.evaluate(xT, yT, batch_size=100)

 8500/10002 [========================>.....] - ETA: 0s[0.64749416913349278, 0.59

738052658136309]

How to interpret the message above? What is the accuracy of the model in % according to the output above? 

Thanks and regards,

Nagabhushan

[Matias Valdenegro]

Assuming your model was compiled with the accuracy metrics, then the first
element in evaluate's return is the loss value, and the second is the
accuracy, so your model has 59% accuracy.

PD: It is not necessary to post questions more than once.

[simply....@gmail.com]

I am getting a similar output, and curious to understand - why the evaluate function does not cover all the datapoints?

In Nagbhushan's output it has covered only 8500/10002 datapoints and in mine it has covered only 32/768 datapoints.

scores = model.evaluate(X,Y)

print("\n%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))

 32/768 [>.............................] - ETA: 0s
acc: 75.52%

Thanks.

[Klemen Grm]

That's probably just an error rendering the progress bar. You can verify it by getting the model predictions and calculating the accuracy separately.