Tree-Based Methods for Statistical Learning in R: A Practical Introduction with Applications in R

Tree-based Methods for Statistical Learning in R  provides a thorough introduction to both individual decision tree  algorithms (Part I) and ensembles thereof (Part II). Part I of the book  brings several different tree algorithms into focus, both conventional  and contemporary. Building a strong foundation for how individual  decision trees work will help readers better understand tree-based  ensembles at a deeper level, which lie at the cutting edge of modern  statistical and machine learning methodology.

The  book follows up most ideas and mathematical concepts with code-based  examples in the R statistical language; with an emphasis on using as few  external packages as possible. For example, users will be exposed to  writing their own random forest and gradient tree boosting functions  using simple for loops and basic tree fitting software (like rpart and party/partykit),  and more. The core chapters also end with a detailed section on  relevant software in both R and other opensource alternatives (e.g.,  Python, Spark, and Julia), and example usage on real data sets. While  the book mostly uses R, it is meant to be equally accessible and useful  to non-R programmers.

Consumers of this book will  have gained a solid foundation (and appreciation) for tree-based methods  and how they can be used to solve practical problems and challenges  data scientists often face in applied work.

Features:

• Thorough  coverage, from the ground up, of tree-based methods (e.g., CART,  conditional inference trees, bagging, boosting, and random forests).
• A companion website containing additional supplementary material and the code to reproduce every example and figure in the book.
• A companion R package, called treemisc,  which contains several data sets and functions used throughout the book  (e.g., there’s an implementation of gradient tree boosting with LAD  loss that shows how to perform the line search step by updating the  terminal node estimates of a fitted rpart tree).
• Interesting  examples that are of practical use; for example, how to construct  partial dependence plots from a fitted model in Spark MLlib (using only  Spark operations), or post-processing tree ensembles via the LASSO to  reduce the number of trees while maintaining, or even improving  performance.