Chapter 33 Further Reading: Introduction to Machine Learning for Business

The resources below are curated for business practitioners — people who want to use ML effectively without becoming research scientists. Each entry includes a note on what it is good for and who should read it.

Books

Conceptual Foundations

"The Hundred-Page Machine Learning Book" — Andriy Burkov The best single-volume technical primer for practitioners who want to understand what the algorithms actually do without drowning in notation. Remarkably concise and honest about limitations. Read this before or alongside Chapter 34.

"Introduction to Statistical Learning" (ISLR) — James, Witten, Hastie, Tibshirani The standard accessible textbook for supervised learning. Available free as a PDF from the authors' website. Chapters 1–4 are directly relevant to this chapter and Chapter 34. The R examples can be mentally translated to Python.

"Data Science for Business" — Provost and Fawcett Focuses specifically on how ML thinking applies to business decision-making. Strong chapters on the business value of prediction, the confusion matrix in business terms, and the economics of model deployment. Does not require coding.

"Weapons of Math Destruction" — Cathy O'Neil Essential reading on what can go wrong when models are deployed at scale without adequate oversight. Keeps you honest about the real-world consequences of algorithmic decision-making. Not a how-to book — a how-not-to book.

Practical Implementation

"Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow" — Aurélien Géron The most comprehensive practical guide to scikit-learn in print. Chapters 1–3 and the scikit-learn sections of later chapters are directly applicable to this book. Dense but excellent.

"Python Machine Learning" — Sebastian Raschka and Vahid Mirjalili Strong on the mathematical intuition behind algorithms, presented accessibly. Good companion to this text if you want to understand why the algorithms work, not just how to use them.

Online Courses

fast.ai "Practical Deep Learning for Coders" Despite the name, the first few lessons cover practical ML fundamentals with a top-down, application-first philosophy. The approach of starting with a working model and then understanding the details aligns well with this textbook.

Google Machine Learning Crash Course Free, browser-based, and covers the fundamentals of ML with TensorFlow examples. Particularly strong on the conceptual sections (overfitting, evaluation, feature engineering). Available at developers.google.com/machine-learning/crash-course.

Kaggle "Intro to Machine Learning" and "Intermediate Machine Learning" Free, interactive courses built around real datasets. The best way to get hands-on practice with scikit-learn workflows outside of this textbook. Kaggle also has excellent competitions for practice.

Papers and Articles

"A Few Useful Things to Know About Machine Learning" — Pedro Domingos (2012) A classic 10-page paper covering the practical realities of applied ML. Covers generalization, overfitting, the curse of dimensionality, and the importance of features. Highly readable. Search the title to find the free PDF.

"Rules of Machine Learning: Best Practices for ML Engineering" — Martin Zinkevich (Google) A list of 43 rules for practical ML development, written for engineers at Google but applicable broadly. Rule 1 is: "Don't be afraid to launch a product without machine learning." Available free on Google's developer site.

"Machine Learning: The High-Interest Credit Card of Technical Debt" — Sculley et al. (Google, 2015) An important paper on the maintenance costs of deployed ML systems. Short, readable, and sobering. Makes the case for preferring simpler, more maintainable models over complex ones.

Tools and Documentation

scikit-learn User Guide (scikit-learn.org/stable/user_guide.html) The official documentation is unusually good. The "Model Selection" and "Metrics and Scoring" sections are directly relevant to this chapter. Read the conceptual sections, not just the API reference.

scikit-learn Examples Gallery Dozens of annotated, runnable examples illustrating specific techniques. Search for "classification" or "cross-validation" to find relevant examples.

pandas Documentation Data preparation is more important than model selection. Strong pandas skills are a prerequisite for effective ML. Review the "Reshaping and Pivot Tables" and "Time Series" sections if you plan to build time-series models.

Podcasts

"Super Data Science" — Jon Krohn Long-form interviews with ML practitioners across a range of industries. Good for understanding how ML is applied in business contexts beyond the textbook.

"Data Skeptic" Critical, rigorous coverage of data science topics. Lives up to its name — often covers the limitations and failure modes of ML techniques, which is exactly what business practitioners need to hear.

"Practical AI" — Changelog Focused on applied AI and ML rather than research. Good for staying current on tools and industry applications without needing to read research papers.

On Critical Thinking About ML Claims

Before accepting any ML result — in a case study, a vendor pitch, or your own model output — apply these questions:

  1. What is the baseline performance? How much better is the ML model?
  2. Was the evaluation performed on data the model was trained on, or truly held-out data?
  3. What is the class distribution? Would a naive majority-class predictor achieve similar accuracy?
  4. How was the training data collected? Could there be selection bias?
  5. Has the model been deployed? What happened when it met real-world data?
  6. Who funded the case study? Does the publisher have a financial interest in positive results?

These questions will serve you better than any algorithm. The most important skill in applied ML is not coding or statistics. It is the discipline to ask uncomfortable questions about your own models and other people's claims.