14. Categorical Data

HOW TO ENCODE CATEGORICAL DATA

Activity

• Looking at the Data.csv, we see that we have two categorical variables which are Country and Purchased

• Their values contain categories

• So we need to encode the text we have over here into numbers

Encoding categorical data in Python

    # Encoding categorical data
    from sklearn.preprocessing import LabelEncoder
    labelEncoder_X = LabelEncoder()
    X[:, 0] = labelEncoder_X.fit_transform(X[:, 0])

• We use the another library LabelEncoder from sklearn and create a class of it

• We then fit and transform the country column

• All of the countries are now encoded in numbers instead

• Now instead replace X[:, 0] country column with the labelEncoder_X.fit_transform(X[:, 0]) to show the countries in text instead

• If you print X on Ipython console, now X contains all number values

• The problem is that you need to somehow make the encoded values make sense

• So we have to prevent the machine learning equations from thinking that Germany is greater than France, and Spain is greater than Germany

• To prevent this we are going to the dummy variables

Dummy Encoding in Python

• Instead of having one column for Country, we are going to have three columns equal to the number of categories countries

• So if the country is France, the France column is going to be one, else zero and likewise for the rest of the columns

    from sklearn.preprocessing import LabelEncoder, OneHotEncoder

    onehotencoder = OneHotEncoder(categorical_features = [0])

• We are going to use OneHotEncoder class to do this, so inspect this class documentation

• Look at the categorical features, and we need to specify the column of the categorical data which is 0

    X = onehotencoder.fit_transform(X).toarray()

• The following method will encode the 0 column of X

• Ctrl Enter to see if the code can run, and print X in Ipython Console

• If you look at it in variable explorer, you can now see the values being encoded in 0s and 1s for Countries column

• Change the formatting of decimas to %.0f to no decimals if your data cannot be seen easily

• Now if you open the dataset, you can see the Countries column being replaced in X by 3 columns

    labelEncoder_Y = LabelEncoder()
    Y = labelEncoder_Y.fit_transform(Y)

• So now we are going to do the same thing for Y, for Purchased column

• Since there's only 2 categories we do not need to onehotencoder to encode more columns

Encoding categorical data in R

    # Encoding categorical data
    dataset$Country = factor(dataset$Country, levels = c('France', 'Spain', 'Germany'), labels = c(1, 2, 3))

• We use a factor method to specify a vector, and c is a vector in R and we creating a vector of 3 elements of Countries

• Labels are used to represent the number associate with that country

• Execute the code and see the difference in output for Data

• Do the same for the Purchased column and you should see the Data being reflected differently for categorical columns of Country and Purchased.