Ames Housing — Categorical Encoding

Demonstrates three categorical encoding techniques on the Ames Housing dataset, each applied to a column chosen because it fits that method best.

Encodings

Technique	Column	Why this column
Label Encoding	Kitchen Qual	Ordinal — natural rank order (Poor < Fair < Typical/Average < Good < Excellent)
One-Hot Encoding	Bldg Type	Nominal, low cardinality (5 categories, no order)
Frequency Encoding	Neighborhood	Nominal, high cardinality (28 categories) — one-hot would create too many sparse columns

Contents

• Ames_Housing_Encodings.ipynb — full notebook: data inspection, all three encodings (each in its own dataframe copy with validation checks), pre-/post-encoding visualizations, and a correlation matrix against SalePrice
• AmesHousing.csv — dataset used

Notebook sections

1. Setup & load data
2. Inspect target columns for missing values
3. Label encoding — Kitchen Qual
4. One-hot encoding — Bldg Type
5. Frequency encoding — Neighborhood
6. Summary table
7. Visualizations — pre-encoding distributions, post-encoding distributions, correlation matrix

Key finding

Kitchen Qual_Label correlates strongly with SalePrice (r ≈ 0.67) — kitchen quality is a meaningful price driver in this dataset. The one-hot Bldg Type columns show negative correlation with each other, which is expected: the categories are mutually exclusive by construction.

Requirements

pandas
scikit-learn
matplotlib
seaborn

Usage

jupyter notebook Ames_Housing_Encodings.ipynb