Data Science from Scratch, 2nd Edition, teaches the mathematics and algorithms behind data analysis through Python implementations. Joel Grus covers linear algebra, statistics, probability, optimization, and machine learning by building core methods directly. It is intended for readers who want to understand data science models as well as use them.
Digital ebook. Checkout sends an order request.
Writing the underlying calculations is the central learning method in this book. Python examples introduce vectors, matrices, statistical summaries, probability, and hypothesis testing before moving into optimization and predictive models. Implementing these pieces makes the relationship between mathematical ideas and program behavior explicit.
Gradient descent leads into machine learning topics including decision trees, random forests, neural networks, clustering, and dimensionality reduction. Natural language processing provides further applications of the earlier statistical concepts. Alongside these models, the text considers collecting, cleaning, and visualizing data. Programmers and developing data scientists can follow the implementations to examine how the main components of an analysis fit together.
Grus teaches data science by making readers build it: vectors, probability, then gradient descent, trees, neural networks and clustering, with text processing and data cleaning along the way. Every method is written in Python, so the arithmetic behind familiar libraries becomes visible rather than assumed.
The organizing idea is that writing the calculation is how you learn it. Linear algebra, statistics and probability come first, because later models lean on them. Optimization is introduced as a general procedure, then applied to predictive methods: decision trees, random forests, neural networks, clustering and dimensionality reduction. Natural language processing shows the same statistical ideas working on text. The closing material turns to the mundane side of analysis, gathering, cleaning and plotting data. Readers work in Python throughout, so model behavior and code behavior stay tied together. It suits programmers and newer data scientists who want the mechanics, not just the API calls.
Vectors, matrices, statistical summaries and probability are introduced as things the reader constructs in code. Because each operation is written out, the link between a symbolic expression and what a program actually computes stays explicit. This groundwork is what later chapters quietly assume.
The text moves from probability to hypothesis testing, treating inference as a procedure with assumptions rather than a verdict machine. Readers see how a test is assembled and what its output does and does not support. The aim is to make statistical claims inspectable.
Gradient descent is presented as a general way of improving a model step by step. Rather than hiding it inside a library, the book derives the update and implements it, so the connection between a loss function and learning becomes concrete. Later models reuse this same machinery.
Trees lead into ensembles, neural networks and then unsupervised methods. Each model is built from the tools already covered, which makes the differences between them a matter of structure rather than magic. Clustering and dimensionality reduction extend the discussion to data without labels.
Natural language processing applies the earlier statistics to words, showing how text can be represented and analyzed. The treatment is practical rather than exhaustive, keeping the focus on how techniques carry over from numerical data to language.
Collecting, cleaning and visualizing data get their own attention, since a model is only as good as the data feeding it. Visualization is treated as a way to inspect and communicate, not decoration. This section grounds the modeling chapters in the reality of messy inputs.
Programmers moving into data work and newer data scientists who want to understand the calculations behind common methods rather than only call them.
Edition, audience, and study-guide details at a glance.
Data Science from Scratch, 2nd Edition, teaches the mathematics and algorithms behind data analysis through Python implementations. Joel Grus covers linear algebra, statistics, probability, optimization, and machine learning by building core methods directly. It is intended for readers who want to understand data science models as well as use them.
This listing is for the 2nd edition. Check the title, ISBN, and book details above before ordering, especially if your course or reading list names a specific edition.
Programmers moving into data work and newer data scientists who want to understand the calculations behind common methods rather than only call them.
Yes. This page includes an original study guide with six thematic sections, key takeaways, and a topic map for structured reading.
More books from this subject area.
