Decision Tree
A tree-structured model that makes predictions through sequential feature-based splits.
A decision tree is a supervised learning model that represents a series of hierarchical decisions as a branching tree structure. Each internal node tests a specific input feature, each branch corresponds to the outcome of that test, and each leaf node holds a final prediction — a class label for classification tasks or a continuous value for regression. This architecture closely mirrors human reasoning processes, making decision trees among the most interpretable models in machine learning.
Training a decision tree involves recursively partitioning the input data by selecting the feature and threshold that best separates the target values at each node. For classification, common splitting criteria include Gini impurity and information gain (based on entropy); for regression, variance reduction is typically used. The algorithm greedily selects the locally optimal split at each step, continuing until a stopping condition is met — such as a maximum tree depth, a minimum number of samples per leaf, or pure node composition. Without constraints, trees tend to overfit the training data, memorizing noise rather than learning generalizable patterns. Pruning techniques, either during growth (pre-pruning) or after (post-pruning), help address this by removing branches that provide little predictive power.
Decision trees gained significant traction in machine learning through Ross Quinlan's ID3 algorithm in the 1980s, followed by C4.5 and C5.0, which added support for continuous attributes, missing values, and more robust pruning. These advances made decision trees practical for real-world datasets. A key strength is their ability to handle both numerical and categorical features without requiring normalization or extensive preprocessing, and their learned structure can be directly inspected and explained to non-technical stakeholders.
While individual decision trees are often outperformed by more complex models, they serve as the foundational building block for powerful ensemble methods such as Random Forests and Gradient Boosted Trees, which combine many trees to dramatically improve predictive accuracy and robustness. Understanding decision trees is therefore essential not only for their direct applications but as a gateway to some of the most effective algorithms in modern machine learning.
