Leading With Data· a Bicycle Guide

Partial book profile

Data Science for Business: What You Need to Know about Data Mining and Data-Analytic Thinking

A conceptual guide that distills the fundamental principles underlying data science so that business people and aspiring data scientists can think data-analytically about extracting useful knowledge from data to improve business decisions.

The full book profile is not available yet.

Get the book →

What it’s about

Data Science for Business is the definitive primer for understanding data science not as a grab-bag of algorithms but as a coherent set of fundamental principles that structure data-analytic thinking. Provost and Fawcett—both seasoned practitioners and researchers—argue that beneath the dizzying array of data mining techniques lies a relatively small set of concepts (treating data as a strategic asset, framing problems with expected value, finding informative attributes, fitting models while controlling overfitting, measuring similarity) that unify the field. Organized around the CRISP data mining process and richly illustrated with real-world business cases—customer churn, targeted marketing, fraud detection, charity solicitation, whiskey recommendation, text mining of news—the book teaches readers to decompose business problems into solvable data science tasks, to evaluate solutions in business terms, and to communicate across the technical/business divide. It is the rare book that equips managers to evaluate data science proposals and equips data scientists to align their work with business value, making both better at extracting competitive advantage from data.

The through-line

Who it’s for
A business professional, manager, investor, or aspiring data scientist who wants to extract competitive advantage and better decisions from their organization's data.
The problem
They have vast amounts of data but lack a principled way to turn it into useful knowledge and better business decisions. They feel intimidated by jargon and algorithms, unsure whether a proposed data science effort is sound or whether they are being misled.
The plan
  1. Learn the small set of fundamental concepts that underlie data science.
  2. Adopt data-analytic thinking and the CRISP process to structure problems.
  3. Decompose business problems into known data mining tasks using the expected value framework.
  4. Evaluate models in business terms, guarding against overfitting and misleading metrics.
  5. Build, nurture, and manage data science capability as a strategic asset.
The payoff
The reader confidently frames business problems data-analytically and decomposes them into solvable tasks. · The reader can evaluate data science proposals, spot flaws, and ask probing questions. · The reader's organization invests wisely in data and data scientists, gaining and sustaining competitive advantage.

See our guide

Related profiles we’ve built

Additional reading

  • The Elements of Statistical Learning: Data Mining, Inference, and Prediction · Hastie, T., Tibshirani, R., & Friedman, J.

    The book refers to this as an excellent technical discussion for readers interested in the details of regularization techniques like ridge regression and the lasso.

  • To explain or to predict? · Shmueli, G.

    The book recommends this paper for readers interested in the deep-felt debate around the different goals of explanatory modeling versus predictive modeling.

  • An introduction to ROC analysis · Fawcett, T.

    Cited as a key reference for the Receiver Operating Characteristics (ROC) graph, a critical tool for model evaluation discussed in detail in the book.

  • Privacy in Context · Nissenbaum, H.

    Recommended as a book that delves deeply into the relationship between privacy and massive databases, for readers who want to think more seriously about the privacy and ethical implications of data science.

  • Matrix factorization techniques for recommender systems · Koren, Y., Bell, R., & Volinsky, C.

    Cited as a key paper written by members of the winning Netflix Challenge team, explaining the latent dimension approach (matrix factorization) for building recommender systems.

  • Classification and Regression Trees · Breiman, L., Friedman, J., Olshen, R., & Stone, C.

    Cited as one of the seminal, popular, and still widely used tree induction systems (CART).

Featured in these guides

Co-cited books