1. Learn
    2. /
  2. कोर्स
    3. /
  3. Recurrent Neural Networks (RNNs) for Language Modeling with Keras
Connected

अभ्यास

Precision-Recall trade-off

When working with classification tasks, the term Precision-Recall trade-off often appears. Where does it comes from?

Usually, the class with higher probability is chosen to assign the document to. But, what if the maximum probability is equal to 0.1? Should you consider that document to belong to this class with only 10% probability?

The answer varies according to problem at hand. It is possible to add a minimum threshold to accept the classification, and by changing the threshold the values of precision and recall move in opposite directions.

The variables y_true and the model model are loaded. Also, if the probability is lower than the threshold, the document will be assigned to DEFAULT_CLASS (chosen to be class 2).

निर्देश

100 XP
  • Use the .predict() method to get the probabilities for each class and store them in the pred_probabilities variable.
  • Accept the maximum probability only if it is greater than or equal to 0.5 and store the results in the y_pred_50 variable.
  • Use the np.argmax() and np.max() functions to do the same for a threshold equal to 0.8.
  • Print the trade_off variable with all the metrics.