How to get class prediction for new entry value in neural networks?

I am new to Nerual Networks model building !! I am doing an multi-class text classification using neural networks.

Steps I have done :

1.Data Cleaning
2.Keras – Text Vectorization for Input Data
3.Dummie values for Label data

I have build the model like this:

model = tf.keras.Sequential([
  layers.Embedding(max_features + 1, embedding_dim),
  layers.Dropout(0.2),
  layers.GlobalAveragePooling1D(),
  layers.Dropout(0.2),
  layers.Dense(10)])

model.summary()

model.compile(loss=losses.BinaryCrossentropy(from_logits=True),
              optimizer='adam',
              metrics=tf.metrics.BinaryAccuracy(threshold=0.0))

epochs = 10
history = model.fit(X_train,Y_train , validation_data=(X_test , Y_test), epochs=epochs)

I am getting 90 accuracy , but the problem is when I want to test with a single text sample , I am getting an array of values like these [-1.975121 , -2.8071122, -2.183652 , …, -2.3590472, -2.214075 ,
-1.9891262] ,but in these which one I should pick ?

And How to do it ??

>Solution :

I am going to guess based on your model shape (layers.Dense(10)) that you’ve 10 classes to classify. In that case, when you do a model.predict you’d get an array of shape (10,) where each element corresponds to a class.

The highest scoring index is the class the model classified. You can get it by:

import numpy as np
predicted_class = np.argmax(<your_models_prediction_array>)

Also, considering it’s a multiclass classification task you should change your model.compile to match >2 (non-binary) classification task:

model.compile(loss='categorical_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])