What is a Decision Tree?
A decision tree is a visual model that helps users make decisions or predictions by splitting data into smaller and smaller groups based on certain rules. An algorithm bases each split on a question or condition, and the structure resembles a branching tree—starting from a single root and expanding into multiple outcomes. Decision trees are popular because they are easy to interpret and mimic the way humans naturally think through problems: step by step, with “if–then” logic.
The purpose of a decision tree is to guide decision-making or classification by breaking a complex question into a series of simpler, binary choices. Users can follow a clear path from the root of the tree to a final decision or predicted outcome. This makes decision trees especially useful when you want transparency in how conclusions are reached.
Some decision trees, especially those used in decision analysis, include extra information such as probabilities and payoffs. The algorithm assigns probabilities to the branches to represent the likelihood of each outcome, while payoffs (or costs) appear at the leaf nodes to show the result of each decision path. These enhanced trees help users calculate the expected value of different choices and identify the most favorable strategy under uncertainty. This format is especially useful in business, economics, and healthcare when evaluating risk or making high-stakes decisions.
Analysts use decision trees widely across many fields. In data science and machine learning, they help classify items or predict values based on input features. Businesses use decision trees to evaluate strategic choices, weighing risks and benefits. In healthcare, they support clinical decision-making, such as diagnosing conditions or choosing treatment plans. In operations and logistics, decision trees help map out workflows or assess risks in complex systems. Their transparency and simplicity make them a practical choice whenever clear, rule-based decisions are needed.
Decision Tree Nodes
A decision tree has three basic types of nodes:
-
Root node: The starting point that represents the entire dataset. It splits based on the most informative question.
-
Decision (or internal) nodes: These represent questions or tests that divide the data into subgroups.
-
Leaf nodes (or terminal nodes): These show the final outcome or decision, such as a predicted class or value.
Algorithms
Several types of algorithms create decision trees by determining the best places to split the data. These methods aim to improve how well the tree classifies or predicts based on the training data. Common methods include:
- CART (Classification and Regression Trees): Uses Gini impurity or mean squared error.
- ID3: Uses information gain to decide splits.
- C4.5: An extension of ID3 that handles both categorical and continuous data.
- CHAID: Uses chi-square tests to split nodes based on statistical significance
Decision Tree Example
Suppose a bank wants to predict whether a customer will default on a loan. The decision tree might use data about income, credit score, and employment status to make this prediction.
This decision tree shows how the model makes predictions based on step-by-step questions. If a customer has an income above $50,000 and a credit score above 700, the tree predicts they will not default. However, if the customer has lower income or lacks employment, the tree leans toward predicting a default. The benefit of using a decision tree here is its clarity—you can see exactly how the decision is made at each step.
« Back to Glossary Index