Polynomial Features

# Import pandas and PolynomialFeatures
import pandas as pd
from sklearn.preprocessing import PolynomialFeatures

# Monthly energy data — two input features
energy_df = pd.DataFrame({
    'month':          ['Jan','Feb','Mar','Apr','May',
                       'Jun','Jul','Aug','Sep','Oct'],
    'avg_temp_c':     [2, 4, 10, 15, 20, 26, 30, 28, 21, 13],
    'household_size': [2, 3, 4, 2, 3, 4, 2, 3, 4, 2],
    'kwh_consumed':   [520,490,310,260,280,410,510,480,290,320]
})

# Select the two input features — exclude month and target
feature_cols = ['avg_temp_c', 'household_size']
X = energy_df[feature_cols]

# degree=2 generates: original terms, squared terms, and interaction term
# include_bias=False removes the constant 1 column (handled by the model intercept)
poly = PolynomialFeatures(degree=2, include_bias=False)

# fit_transform learns the feature names and generates all polynomial terms
X_poly = poly.fit_transform(X)

# get_feature_names_out shows exactly which column maps to which term
poly_names = poly.get_feature_names_out(feature_cols)
poly_df = pd.DataFrame(X_poly, columns=poly_names)

# Print the feature names to understand what was generated
print("Generated polynomial features:")
for name in poly_names:
    print(f"  {name}")

print()
print(poly_df.to_string(index=False))

# Import pandas and PolynomialFeatures
import pandas as pd
from sklearn.preprocessing import PolynomialFeatures

# Sales campaign data across regional markets
sales_df = pd.DataFrame({
    'region':       ['North','South','East','West','North',
                     'South','East','West','North','South'],
    'ad_spend':     [50,80,120,40,90,110,60,70,100,55],
    'price_discount':[5,10,15,0,20,10,5,15,20,0],
    'market_size':  [200,350,500,150,280,420,190,310,460,175]
})

# interaction_only=True generates ONLY cross-product terms, no squared terms
# degree=2 with interaction_only gives: x*z, x*w, z*w for three features
poly_interact = PolynomialFeatures(degree=2, interaction_only=True, include_bias=False)

# Fit and transform the three numerical features
feature_cols = ['ad_spend', 'price_discount', 'market_size']
X_interact = poly_interact.fit_transform(sales_df[feature_cols])

# Get the generated column names
interact_names = poly_interact.get_feature_names_out(feature_cols)
interact_df = pd.DataFrame(X_interact, columns=interact_names, dtype=int)

# Print feature names and the first five rows
print("Interaction-only features generated:")
for name in interact_names:
    print(f"  {name}")
print()
print(interact_df.to_string(index=False))

# Import all tools needed
import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler, PolynomialFeatures
from sklearn.pipeline import Pipeline
from sklearn.model_selection import train_test_split

# Re-use the energy dataset from earlier in this lesson
energy_df = pd.DataFrame({
    'avg_temp_c':    [2, 4, 10, 15, 20, 26, 30, 28, 21, 13],
    'household_size':[2, 3, 4, 2, 3, 4, 2, 3, 4, 2],
    'kwh_consumed':  [520,490,310,260,280,410,510,480,290,320]
})

# Separate features and target, then split
X = energy_df[['avg_temp_c', 'household_size']]
y = energy_df['kwh_consumed']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Build a Pipeline: scale first, then generate polynomial terms
# Order matters: scale raw features before squaring so the terms stay well-behaved
poly_pipeline = Pipeline([
    ('scaler', StandardScaler()),
    ('poly',   PolynomialFeatures(degree=2, include_bias=False))
])

# Fit on training data, transform both sets
X_train_poly = poly_pipeline.fit_transform(X_train)
X_test_poly  = poly_pipeline.transform(X_test)

# Retrieve the final column names from the poly step
feature_names = poly_pipeline.named_steps['poly'].get_feature_names_out(X.columns)

# Print shape and first few rows of the transformed training set
print(f"Input shape:  {X_train.shape}")
print(f"Output shape: {X_train_poly.shape}")
print(f"Features:     {feature_names.tolist()}")
print()
result = pd.DataFrame(X_train_poly.round(3), columns=feature_names)
print(result.to_string(index=False))