Machine Learning Algorithms - Second Edition

Managing categorical data

In many classification problems, the target dataset is made up of categorical labels that cannot immediately be processed by every algorithm. An encoding is needed, and scikit-learn offers at least two valid options. Let's consider a very small dataset made of 10 categorical samples with 2 features each:

import numpy as npX = np.random.uniform(0.0, 1.0, size=(10, 2))Y = np.random.choice(('Male', 'Female'), size=(10))print(X[0])array([ 0.8236887 ,  0.11975305])print(Y[0])'Male'

The first option is to use the LabelEncoder class, which adopts a dictionary-oriented approach, associating to each category label a progressive integer number, that is, an index of an instance array called classes_:

from sklearn.preprocessing ...