One hot encoding increases sizes of target data - machine-learning

Question

i have multiclass data, label or y columns contains following data :
print(y.unique())
[5 6 7 4 8 3 9]
in this case number of class is equal to 7(when modelling of deep learning), but when i do one hot encoding like this :
import keras
from keras.utils import np_utils
y_train =np_utils.to_categorical(y_train)
y_test =np_utils.to_categorical(y_test)
dimension increased to 10
print(y_train.shape) : (4547, 10)
maybe because we have numbers up to 9 and (0,1,2) is also included(in fact it is not represented in original data), how can i fix this issue?

Answer 1

The function tf.keras.utils.to_categorical requires the inputs to be "integers from 0 to num_classes" (see the documentation). You have a set of labels {3, 4, 5, 6, 7, 8, 9}. That is a total of seven labels, which start at the value 3. To transform this to a set of labels in [0, 7), one can subtract 3 from each label.
y_ints = y - 3
The result can be passed to tf.keras.utils.to_categorical.
import numpy as np
import tensorflow as tf
y = np.array([3, 4, 5, 6, 7, 8, 9])
y_ints = y - 3 # [0, 1, 2, 3, 4, 5, 6]
tf.keras.utils.to_categorical(y_ints)
and output is
array([[1., 0., 0., 0., 0., 0., 0.],
[0., 1., 0., 0., 0., 0., 0.],
[0., 0., 1., 0., 0., 0., 0.],
[0., 0., 0., 1., 0., 0., 0.],
[0., 0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 0., 1., 0.],
[0., 0., 0., 0., 0., 0., 1.]], dtype=float32)
Another option is to use scikit-learn's extensive preprocessing methods, in particular sklearn.preprocessing.OneHotEncoder.
import numpy as np
from sklearn.preprocessing import OneHotEncoder
y = np.array([3, 4, 5, 6, 7, 8, 9])
y = y.reshape(-1, 1) # reshape to (n_samples, n_labels).
encoder = OneHotEncoder(sparse=False, dtype="float32")
encoder.fit_transform(y)
The output is
array([[1., 0., 0., 0., 0., 0., 0.],
[0., 1., 0., 0., 0., 0., 0.],
[0., 0., 1., 0., 0., 0., 0.],
[0., 0., 0., 1., 0., 0., 0.],
[0., 0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 0., 1., 0.],
[0., 0., 0., 0., 0., 0., 1.]], dtype=float32)