Transforming categorical variables
HR Analytics: Predicting Employee Churn in Python
Hrant Davtyan
Assistant Professor of Data Science American University of Armenia
Types of categorical variables
- Ordinal - variables with two or more categories that can be ranked or ordered
- Our example: salary
- Values: low, medium, high
- Nominal - variables with two or more categories with do not have an intrinsic order
- Our example: department
- Values: sales, accounting, hr, technical, support, management, IT, product_mng, marketing, RandD
Encoding categories (salary)
# Change the type of the "salary" column to categorical
data.salary = data.salary.astype('category')
# Provide the correct order of categories
data.salary = data.salary.cat.reorder_categories(['low',
'medium',
'high'])
# Encode categories with integer values
data.salary = data.salary.cat.codes
| Old values | New values |
|---|---|
| low | 0 |
| medium | 1 |
| high | 2 |
Getting dummies
# Get dummies and save them inside a new DataFrame
departments = pd.get_dummies(data.department)
Example output:
IT RandD accounting hr management marketing product_mng sales support technical
0 0 0 0 0 0 0 0 0 0 1
Dummy trap
departments.head()
IT RandD accounting hr management marketing product_mng sales support technical
0 0 0 0 0 0 0 0 0 0 1
departments = departments.drop("technical", axis = 1)
departments.head()
IT RandD accounting hr management marketing product_mng sales support
0 0 0 0 0 0 0 0 0 0
Let's practice!
HR Analytics: Predicting Employee Churn in Python
