Ciaren

Encode Categories — encodeCategories

Turn the values of selected categorical fields into flags or codes.

Use cases

  • Inspect one-hot columns for a discrete category.
  • Explore ordinal codes before choosing a model preprocessing strategy.

What it does

With onehot, pandas get_dummies replaces the selected columns with category flags; drop_first optionally omits the first category per column. Other columns and rows remain. With ordinal, an OrdinalEncoder fits string representations of the selected values and replaces those columns with numeric codes.

The node fits on each incoming frame and emits only out, not a saved encoder. Ordinal codes and one-hot schemas can change when the observed categories change. Use the train node's bundled categorical preprocessing for a fitted encoder that will be reused at prediction time.

Configuration

Config keyTypeRequiredDescription
columnsstring[]YesNon-empty list of existing columns to treat as categories
methodstringNoonehot (default) or ordinal
drop_firstbooleanNoDefault false; used only by onehot

Generated Python code

Download the customer sample CSV and save it as customers.csv beside the script. These are complete exports from File Input → this node → File Output; additional preparation steps are stated below.

Here the discrete support_calls count is treated as a category with onehot and drop_first: false. The scripts write encode-categories.csv: support_calls is replaced by support_calls_<value> flags, with one active flag per row. All other fields, including churn, remain unchanged.

The node configuration is:

json
{
  "method": "onehot",
  "columns": ["support_calls"],
  "drop_first": false
}

Both flow engines run this ML step through pandas/scikit-learn. The Polars script includes the actual conversion into pandas and back into Polars.

pandas

python
import pandas as pd

df_customers = pd.read_csv('customers.csv')
df_customers = pd.get_dummies(df_customers, columns=['support_calls'], drop_first=False)
df_customers.to_csv('encode-categories.csv', index=False)

Polars

python
import polars as pl
import pandas as pd

df_customers = pl.read_csv('customers.csv')
_eager_1 = df_customers.to_pandas()
_eager_2 = pd.get_dummies(_eager_1, columns=['support_calls'], drop_first=False)
df_1 = pl.from_pandas(_eager_2)
df_1.write_csv('encode-categories.csv')

Tips & common mistakes

  • A selected column is treated as categorical even when its values are numeric. Choose this intentionally.
  • High-cardinality IDs can create many dummy columns; do not encode an ID merely because it is present.
  • Do not fit separate encoders on train and test data or assume independently fitted ordinal codes have the same meaning.
  • Ordinal values are stringified before fitting; fill missing values first if you need an explicit missing-value category.

See also