Feature Engineering Basics

What is Feature Engineering?

Feature engineering is the process of transforming raw data into meaningful inputs that help machine learning models perform better. Good features often matter more than fancy algorithms.

• Handle missing values and inconsistent formats.
• Scale numeric features to similar ranges.
• Encode categorical variables into numbers.
• Create new features from existing ones.

Scaling Numeric Features

Many algorithms (e.g., linear models, KNN, neural networks) work better when features are on similar scales.

import pandas as pd
from sklearn.preprocessing import StandardScaler, MinMaxScaler

df = pd.DataFrame({
    "age": [18, 25, 40, 60],
    "salary": [20000, 35000, 60000, 120000]
})

print("Original data:")
print(df)

# Standardization: mean=0, std=1
std_scaler = StandardScaler()
df_std = pd.DataFrame(
    std_scaler.fit_transform(df),
    columns=df.columns
)

print("\nStandardized features:")
print(df_std.round(2))

# Min-max scaling: range [0, 1]
mm_scaler = MinMaxScaler()
df_mm = pd.DataFrame(
    mm_scaler.fit_transform(df),
    columns=df.columns
)

print("\nMin-max scaled features:")
print(df_mm.round(2))

Encoding Categorical Variables

Models work with numbers, not text. We convert categories to numeric form using techniques like one-hot encoding.

import pandas as pd

df = pd.DataFrame({
    "city": ["Delhi", "Mumbai", "Delhi", "Chennai"],
    "gender": ["M", "F", "F", "M"]
})

print("Original:")
print(df)

# One-hot encode categorical variables
df_encoded = pd.get_dummies(df, columns=["city", "gender"], drop_first=True)

print("\nEncoded:")
print(df_encoded)

Creating New Features

Sometimes you can combine or transform existing columns to capture useful patterns.

import pandas as pd

df = pd.DataFrame({
    "years_experience": [1, 3, 5, 10],
    "salary": [30000, 45000, 70000, 120000]
})

# Create a new feature
df["salary_per_year"] = df["salary"] / df["years_experience"]

print(df)