SQL for Machine Learning Data Preparation : 30 Exercises on Feature Engineering, Normalization, and Queries

Supercharge Your Machine Learning Pipelines: Master SQL for Data Preparation with 30 Hands-On Exercises!
Are you a data scientist or ML engineer struggling to wrangle massive datasets efficiently before modeling? Unlock the full potential of SQL as your secret weapon for seamless data prep in SQL for Machine Learning Data Preparation: 30 Exercises on Feature Engineering, Normalization, and Queries by Cyrus Laban. This practical, exercise-driven guide demystifies how to use SQL to extract, transform, and optimize data directly in the database—saving time, reducing overhead, and scaling effortlessly for real-world ML projects in industries like finance, e-commerce, healthcare, and more.
Perfect for intermediate SQL users with basic ML knowledge, this book bridges the gap between database querying and machine learning workflows. Through 30 progressively challenging exercises using sample datasets (like sales transactions, customer demographics, and financial records), you'll build skills from scratch, applying concepts to practical scenarios such as fraud detection, churn prediction, and recommendation systems. No more switching between tools—learn to handle everything in SQL for reproducible, high-performance results.
What You'll Master Inside:

• Introduction and Fundamentals (Chapters 1-2): Dive into SQL's role in ML data prep, then tackle basic queries for extraction, joins, aggregations, and handling imbalanced data. Exercises 1-5 include extracting features from sales data, joining customer/product tables, aggregating metrics, filtering outliers, and balancing classes via sampling.
• Advanced Techniques (Chapter 3): Level up with subqueries, CTEs, window functions, pivoting, and regex for complex datasets. Exercises 6-10 cover multi-step fraud detection queries, time-series trend analysis, ranking user engagement, pivoting categorical data, and text feature extraction.
• Feature Engineering Mastery (Chapter 4): Transform raw data into powerful features using calculations, binning, interactions, and encoding. Exercises 11-20 guide you through creating ratios (e.g., debt-to-income), date-based features, one-hot encoding, aggregating time windows, interaction terms, binning ages/prices, handling missing values, polynomial features, text tokenization, and regional averages.
• Normalization and Scaling (Chapter 5): Ensure model-ready data with min-max, z-score, log transformations, and robust scaling. Exercises 21-25 focus on scaling housing prices, standardizing customer spends, log-normalizing skewed distributions, robust scaling for outliers, and group-based z-scores.
• Integration and Best Practices (Chapters 6-7): Seamlessly connect SQL to Python/ML tools, optimize queries for big data, and apply ethical considerations. Exercises 26-30 include pipeline queries for image metadata, exporting datasets to CSV, automating feature prep with views, optimizing massive queries, and end-to-end churn prediction.
• Real-World Applications and Resources: Explore case studies, performance tips, and cloud integrations (e.g., BigQuery, Snowflake). Plus, appendices with sample datasets/schemas, detailed exercise solutions, an SQL cheat sheet tailored for ML, and a glossary for quick reference.

Forget theoretical fluff—this is a hands-on workbook that has you querying and engineering from page one, building a portfolio of ML-ready datasets. By the end, you'll confidently prepare data at scale, boosting model accuracy and efficiency. First edition, published in 2025—stay at the forefront of data-driven AI!