Split
A decision-tree split is a rule that partitions observations at a node into subsets according to a feature and a condition on its value.
Statistics Every Programmer Needs
A split is a decision rule that partitions observations at a decision-tree node according to a feature and condition.
- How it is used
- Partitions observations into increasingly specific decision paths for classification or regression predictions.
- How deep it goes
- Explains thresholds, node traversal, root selection, and pruning, but does not develop impurity or pruning mathematics.
- Examples
- Uses TO_GO, PLAY_TYPE, and SCORE_DIFFERENTIAL thresholds to classify fourth-down conversion attempts.
- What you can do after
- Evaluate tree predictions by tracing feature conditions from the root through internal nodes to a terminal leaf.
Grokking Machine Learning, Second Edition
A split is a decision-tree operation that partitions examples at a node into subsets according to a feature-based condition.
- How it is used
- Builds decision trees by repeatedly partitioning data into more classifiable groups.
- How deep it goes
- Explains candidate thresholds, evaluation metrics, purity-based stopping, and split-controlling hyperparameters.
- Examples
- Uses app recommendations split by age and platform, plus student-admission tree settings.
- What you can do after
- Build it from scratch by generating feature questions, evaluating partitions, and continuing until groups become leaves.
Read how Grokking Machine Learning, Second Edition explains split →
Statistics Slam Dunk
A tree split is a rule that partitions observations into child regions according to the value of a predictor.
- How it is used
- Partitions predictor space into regions that support regression-tree predictions.
- How deep it goes
- Explains regional partitioning, split importance, thresholds, and terminal nodes, but not the algorithm selecting each split.
- Examples
- Uses screen_assists_pts greater than 26.05 and a visualized ten-terminal-node wins tree.
- What you can do after
- Evaluate/understand how predictor thresholds divide observations and lead to regional predictions.
Timeless Algorithms, The Foundational Ideas
A decision-tree split partitions observations at a node according to a feature-based rule.
- How it is used
- Partitions observations at tree nodes to separate class labels or reduce variation in a numerical outcome.
- How deep it goes
- Explains split criteria and greedy selection, then traces how early choices can alter later branches.
- What you can do after
- Understand how local split choices affect tree structure and the paths observations follow.
Read how Timeless Algorithms, The Foundational Ideas explains split →
What to know first
- branchA branch is a nonterminal part of a tree that extends from a node toward one or more child nodes, representing a subdivision or decision outcome.
- decision treeA decision tree is a tree-structured model or representation in which successive tests or choices lead through branches to outcomes, predictions, or actions.
- entropyEntropy is an information-theoretic measure of uncertainty in a probability distribution, defined as the expected value of the negative logarithm of outcome probability.
- inputAn input is information, a value, or an action supplied to a system, program, or computation for processing or to influence its behavior.
Related concepts
- decision treepart of · A decision tree is a tree-structured model or representation in which successive tests or choices lead through branches to outcomes, predictions, or actions.
- accuracymeasured by · Accuracy is the degree to which a result agrees with a specified truth, reference, or criterion, commonly measured as the proportion of evaluated predictions that are correct.
- thresholduses · A threshold is a boundary value used to determine whether a measured quantity meets a criterion or changes a classification, decision, or action.
- decision nodeproduces · A decision node is a tree node where a decision, test, or action selection determines which successor branch is followed.
- weighted average Gini impuritymeasured by · Weighted average Gini impurity is the sample-size-weighted average of the Gini impurities of the subsets produced by a decision-tree split.
Questions about split
What is split?
A decision-tree split is a rule that partitions observations at a node into subsets according to a feature and a condition on its value.
Which Manning books explain split?
Statistics Every Programmer Needs, Grokking Machine Learning, Second Edition, Statistics Slam Dunk and Timeless Algorithms, The Foundational Ideas.
What should I know before learning split?
Manning books that teach split usually introduce branch, decision tree, entropy and input first.
What is split part of?
Split is part of decision tree. A decision tree is a tree-structured model or representation in which successive tests or choices lead through branches to outcomes, predictions, or actions.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.



