One-Hot encoding implementation with scikit-learn and dummy variables method

08 Apr, 2022 • 4 min read

One-Hot encoding is one of the approaches of feature encoding. It converts variables that takes multiple values into binary variables (0, 1) for each category. This creates several new variables.

As an example, let's assume a column of your dataset contains a feature called colour. Some of the variables in this feature are blue, green and red. Now what one-hot encoding does is to create new binary variables for each colour, i.e. colour_blue(0, 1), colour_green(0, 1) and colour_red(0, 1).

Create sample dataset

Let's create a small dataset with just 5 features

Create a list of features to be encoded

We want to encode the string categoricals only, so we filter the desired columns.

Method 1: One-Hot encoding using dummy variables

Pandas comes with a built-in 'get_dummies' function which is easy to use for this type of encoding.

Observation

Notice that new variables have been created for every column in the string categorical list, and have been assigned binary values (0, 1)

Method 2: Using the scikit-learn library

We'll be using the OneHotEncoder and LabelEncoder modules from sk-learn.

Write a for loop that performs the encoding functions on each of the columns in the one_hot_encode_cols. Start the loop by 'integer encoding' the string categories using the label encoder.