AI Logic and Robotics Workshops for Kids in Nonthaburi

1. Why Encode Categories

ML models train on numbers, not raw text. Features such as city, gender, or size must be converted before fitting a model. Encoding choice affects accuracy and interpretability.

2. Nominal vs Ordinal

• Nominal (unordered): City, gender, color → One-Hot Encoding
• Ordinal (ordered): Small / Medium / Large, Low / Med / High → Ordinal Encoding

One-hot creates binary columns (exactly one True per row for that feature). Ordinal maps labels to ordered integers in one column.

3. Dimensionality and Rare Levels

High cardinality one-hot expands the feature space (curse of dimensionality). Group rare levels into an Others category before encoding, then inspect with .head() / .info().

4. One-Hot with Pandas

import pandas as pd

data = {
    "City": ["Lahore", "Karachi", "Islamabad", "Lahore", "Karachi"],
    "Segment": ["A", "B", "C", "A", "B"],
    "Sales": [1500, 2200, 1800, 1600, 2100]
}
df = pd.DataFrame(data)

df_encoded = pd.get_dummies(
    df,
    columns=["City", "Segment"],
    drop_first=True
)
print(df_encoded.head())

drop_first=True drops one level per feature to reduce multicollinearity; all-False means the baseline category.

5. Ordinal with Scikit-Learn

import pandas as pd
from sklearn.preprocessing import OrdinalEncoder

df_size = pd.DataFrame({
    "Size": ["Small", "Medium", "Large", "Medium", "Small"]
})
encoder = OrdinalEncoder(categories=[["Small", "Medium", "Large"]])
df_size["Size_Encoded"] = encoder.fit_transform(df_size[["Size"]])
print(df_size)

Result: Small → 0, Medium → 1, Large → 2 in a single numeric column.

6. Quick Comparison

Aspect	One-Hot	Ordinal
Data type	Nominal	Ordinal
Tool	pd.get_dummies	OrdinalEncoder
Output	Many binary columns	One integer column
Risk	High dimensionality	Fake order on unordered data