One-hot encoding in Python
Educative Answers TeamMost of the existing machine learning algorithms cannot be executed on categorical data. Instead, the categorical data needs to first be converted to numerical data. One-hot encoding is one of the techniques used to perform this conversion. This method is mostly used when deep learning techniques are to be applied to sequential classification problems.
One-hot encoding is essentially the representation of categorical variables as binary vectors. These categorical values are first mapped to integer values. Each integer value is then represented as a binary vector that is all 0s (except the index of the integer which is marked as 1).
Manual one-hot encoding
Have a look at the example below which manually converts the categorical list of colors to a numerical list using one-hot encoding:
import numpy as np
### Categorical data to be converted to numeric data
colors = ["red", "green", "yellow", "red", "blue"]
### Universal list of colors
total_colors = ["red", "green", "blue", "black", "yellow"]
### map each color to an integer
mapping = {}
for x in range(len(total_colors)):
mapping[total_colors[x]] = x
one_hot_encode = []
for c in colors:
arr = list(np.zeros(len(total_colors), dtype = int))
arr[mapping[c]] = 1
one_hot_encode.append(arr)
print(one_hot_encode)One-hot encoding using scikit-learn
Take a look at the example below. It uses the scikit-learn library to perform one-hot encoding:
from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import OneHotEncoder ### Categorical data to be converted to numeric data colors = (["red", "green", "yellow", "red", "blue"]) ### integer mapping using LabelEncoder label_encoder = LabelEncoder() integer_encoded = label_encoder.fit_transform(colors) print(integer_encoded) integer_encoded = integer_encoded.reshape(len(integer_encoded), 1) ### One hot encoding onehot_encoder = OneHotEncoder(sparse=False) onehot_encoded = onehot_encoder.fit_transform(integer_encoded) print(onehot_encoded)
RELATED TAGS
View all Courses
