CODEWITHSUNDEEP
python-3.xtensorflowkerasnlp
PeakyBlinder
PeakyBlinder

Reputation: 1117

How to use model.predict in keras?

I am using keras model.predict after training my model for a sentence classification task. My code is

import numpy as np
model = Sequential()
l = ['Hello this is police department', 'hello this is 911 emergency']
tokenizer = Tokenizer()
tokenizer.fit_on_texts(l)
X = tokenizer.texts_to_sequences(l)
X = np.array(X)
a = model.predict(X)
print(a)

But the output seems to be an array,

[[1. 2. 3. 4. 5.]
 [1. 2. 3. 6. 7.]]

I am working on a sentence classification task with 2 labels. So I wanted to predict these sentences as 0 or 1. But instead getting a numpy array. How do I code such that it predicts one of these two labels?

Upvotes: 0

Views: 8280

Answers (1)

Marco Cerliani
Marco Cerliani

Reputation: 22031

add some layer to your model. to get probabilities in [0,1] use sigmoid as last activation

from sklearn.preprocessing import LabelEncoder

maxlen = 10

X_train = ['Hello this is police department', 
     'hello this is 911 emergency',
     'asdsa sadasd',
     'asnxas asxkx',
     'kas',
     'jwxxxx']
y_train = ['positive','negative','positive','negative','positive','negative']

label_enc = LabelEncoder()
label_enc.fit(y_train)

tokenizer = tf.keras.preprocessing.text.Tokenizer()
tokenizer.fit_on_texts(X_train)

X_train = tokenizer.texts_to_sequences(X_train)
X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train, maxlen=maxlen)

y_train = label_enc.transform(y_train)

model = Sequential()
model.add(Dense(1, activation='sigmoid', input_shape=(maxlen,)))
model.compile('adam', 'binary_crossentropy')
model.fit(X_train,y_train, epochs=3)


### PREDICT NEW UNSEEN DATA ###

X_test = ['hello hSDAS', '911 oaoad']

X_test = tokenizer.texts_to_sequences(X_test)
X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test, maxlen=maxlen)

a = (model.predict(X_test)>0.5).astype(int).ravel()
print(a)

reverse_pred = label_enc.inverse_transform(a.ravel())
print(reverse_pred)

Upvotes: 1

Related Questions