CODEWITHSUNDEEP
neural-networkmeasureprecision-recall
alyssaeliyah
alyssaeliyah

Reputation: 2244

Evaluation Measure Neural Network Model Performance

I want to create a neural network model that classifies an input into four classes: Class A, B, C and D. Each Class can be either 0 and 1. What measures shall I use to evaluate my model. Can I use the Precision , Recall, F-measure measures to evaluate the model?

Upvotes: 0

Views: 292

Answers (1)

Parthasarathy Subburaj
Parthasarathy Subburaj

Reputation: 4264

It depends on the data you have,

  1. If your data is quite balanced (i.e. equal representation from all the classes) then you can use Accuracy as a measure.
  2. But if your data is suffering from class imbalance (skewed class distributions) then you should consider class wise precision, recall and F1 score. I recommend f1 score since it considers both precision and recall.
  3. Else you can plot the ROC curve and get the area under ROC.

Sklearn have packages for doing this, you could refer to classification report, Area under ROC .

Upvotes: 1

Related Questions