CODEWITHSUNDEEP
machine-learningkerasneural-network
Enrico Detoma
Enrico Detoma

Reputation: 3179

How to evaluate Keras model precision for only some of the classes

I have the following Keras model, whose output has 3 classes (0, 1, 2):

    model = Sequential()
    model.add(LSTM(100, input_shape=(n_time_steps,n_features)))
    model.add(Dropout(0.5))
    model.add(Dense(100, activation='relu'))
    model.add(Dense(n_classes, activation='softmax'))
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

I'm interested in evaluating the model for precision of classes 1 and 2, not 0, i.e. the model is good if it has the smallest number of false positives in classes 1 and 2, while I don't really care about class 0.

How can I write such metrics in Keras?

Upvotes: 0

Views: 200

Answers (1)

Daniel Möller
Daniel Möller

Reputation: 86620

Create any function taking the ground truth and the predicted results and calculate your metrics:

def false_positives(y_true, y_pred):
    negatives = 1 - y_true   
    y_pred = K.cast(K.greater(y_pred, 0.5), K.floatx()) #round to 0 or 1
    #if you don't round y_pred, it might even serve as a loss function

    falsePositives = y_pred * negatives
    falsePositives1 = falsePositives[:,1]
    falsePositives2 = falsePositives[:,2]

    return something

Use metrics = [false_positives], or metrics=['accuracy', false_positives_1, false_positives_2], etc.

Upvotes: 1

Related Questions