ML Chapter 4 | Data Preprocessing, Cleaning, Encoding & Feature Scaling Explained

Chapter 4: Data Preprocessing & Feature Scaling

Data preprocessing is one of the most important steps in Machine Learning.
Raw data often contains missing values, inconsistent formats, outliers, and mixed datatypes.
Without proper cleaning and preprocessing, even the best algorithms will produce poor results.

In this chapter, we cover everything you need to prepare high-quality data for ML models — including missing values, encoding categorical features, outlier detection, normalization, standardization, and feature engineering.

1. Understanding Data Preprocessing

Data preprocessing refers to preparing raw data into a clean, structured form so that ML algorithms can understand it.

• Handling missing values
• Removing duplicates
• Fixing data types
• Encoding categorical columns
• Scaling and normalization
• Handling outliers

2. Handling Missing Values

✔ Types of Missing Data

• MCAR – Missing Completely at Random
• MAR – Missing at Random
• MNAR – Missing Not at Random

✔ Strategies to Handle Missing Values

• Drop missing rows/columns
• Fill with mean/median/mode
• Use model-based imputation

Python Example (Fill Missing Values)

import pandas as pd

df = pd.DataFrame({
    'age': [20, 25, None, 30],
    'income': [30000, None, 40000, 50000]
})

df['age'].fillna(df['age'].mean(), inplace=True)
df['income'].fillna(df['income'].median(), inplace=True)

print(df)

3. Encoding Categorical Variables

Machine learning models work only with numbers — not text.
Categorical values must be converted into numeric format.

✔ One-Hot Encoding

• Creates new column for each category
• Used when categories have no order

import pandas as pd

data = {'City': ['Delhi', 'Mumbai', 'Delhi', 'Chennai']}
df = pd.DataFrame(data)

encoded = pd.get_dummies(df, columns=['City'])
print(encoded)

✔ Label Encoding

• Assigns integer value to each category
• Used when categories have natural order

from sklearn.preprocessing import LabelEncoder

le = LabelEncoder()
df['City'] = le.fit_transform(df['City'])

print(df)

4. Outlier Detection & Handling

Outliers are extremely high or low values that distort model performance.

✔ Detecting Outliers Using IQR

import numpy as np
import pandas as pd

df = pd.DataFrame({'salary': [1000, 1200, 1100, 90000]})

Q1 = df['salary'].quantile(0.25)
Q3 = df['salary'].quantile(0.75)
IQR = Q3 - Q1

lower = Q1 - 1.5 * IQR
upper = Q3 + 1.5 * IQR

print(df[(df['salary'] < lower) | (df['salary'] > upper)])

✔ How to Handle Outliers?

• Remove them
• Cap them using percentile limits
• Apply log/sqrt transformations
• Use robust models (Random Forest, Tree models)

5. Feature Scaling

Feature scaling ensures that all numerical features contribute equally to the model.
It is crucial for algorithms where distance or gradient is involved such as:

• KNN
• SVM
• K-Means
• Neural Networks
• Gradient-based models

✔ Standardization (Z-Score Scaling)

Transforms data to have mean = 0 and standard deviation = 1.

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
scaled = scaler.fit_transform([[10],[20],[30]])

print(scaled)

✔ Normalization (Min-Max Scaling)

Scales values between 0 and 1.

from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()
scaled = scaler.fit_transform([[10],[20],[30]])

print(scaled)

Standardization vs Normalization

6. Feature Engineering Basics

• Extracting new features from date/time
• Binning numeric values (age groups)
• Combining related features
• Creating polynomial features

Python Example: Creating New Features

df['age_group'] = pd.cut(df['age'], bins=[0,20,40,60,100],
                         labels=['Teen','Adult','Mid-age','Senior'])

7. Complete Preprocessing Pipeline

A full preprocessing pipeline automates the entire cleaning process.

from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer

numeric_features = ['age', 'salary']
categorical_features = ['city']

numeric_pipeline = Pipeline(steps=[
    ('imputer', SimpleImputer(strategy='median')),
    ('scaler', StandardScaler())
])

cat_pipeline = Pipeline(steps=[
    ('imputer', SimpleImputer(strategy='most_frequent')),
    ('encoder', OneHotEncoder(handle_unknown='ignore'))
])

preprocess = ColumnTransformer(
    transformers=[
        ('num', numeric_pipeline, numeric_features),
        ('cat', cat_pipeline, categorical_features)
])

Conclusion

Data preprocessing is the foundation of building successful Machine Learning models.
Clean data → Better accuracy, faster training, and more reliable predictions.

In the next chapter, we will learn about Train-Test Split & Cross-Validation — essential techniques for evaluating model performance.