By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.

Decision Tree Analysis

Decision Tree Analysis Definition

Decision tree analysis is the process of drawing a decision tree, which is a graphic representation of various alternative solutions that are available to solve a given problem, in order to determine the most effective courses of action. Decision trees are comprised of nodes and branches - nodes represent a test on an attribute and branches represent potential alternative outcomes.

Decision tree analysis header image
Image from Lucidspark

FAQs

What is Decision Tree Analysis?

A decision tree is a tree-like model that acts as a decision support tool, visually displaying decisions and their potential outcomes, consequences, and costs. From there, the “branches” can easily be evaluated and compared in order to select the best courses of action.

Decision tree analysis is helpful for solving problems, revealing potential opportunities, and making complex decisions regarding cost management, operations management, organization strategies, project selection, and production methods.

Drawing a decision tree diagram starts from left to right and consists of “burst” nodes that split into different paths. Nodes are categorized as Root nodes, which compiles the whole sample and is then split into multiple sets; Decision nodes, typically represented by squares, are sub-nodes that diverge into further possibilities; and the Terminal node, typically represented by triangles, is the final node that shows the final outcome that cannot be further categorized.

Branches, or lines, represent the various available alternatives, and sub-nodes can be eliminated via Pruning. Decision trees can be hand-drawn or created with the use of decision tree software. Analysis can be performed manually, via decision tree analysis in R, or via automated software.

Five Steps of Decision Tree Analysis

The steps in decision tree analysis consist of:

  1. Define the problem area for which decision making is necessary. 
  2. Draw a decision tree with all possible solutions and their consequences.
  3. Input relevant variables with their respective probability values.
  4. Determine and allocate payoffs for each possible outcome. 
  5. Calculate the Expected Monetary Value for every chance node in order to determine which solution is expected to provide the most value. Circles represent chance nodes in a tree diagram.


Popular applications include: decision tree analysis in risk management, decision tree analysis in healthcare, decision tree analysis in capital budgeting, decision tree business analysis, and decision tree analysis in finance.

Advantages and Disadvantages of Decision Tree Analysis

There are risks and rewards associated with the process of decision tree analysis. The advantages of decision tree analysis include: simple and easy to interpret decision trees; valuable without requiring large amounts of hard data; helps decision makers ascertain best, worst, and expected results for various scenarios; and can be combined with various decision techniques.

When using decision tree analysis, there may also be some disadvantages. Disadvantages include: uncertain values can lead to complex calculations and uncertain outcomes; decision trees are unstable, and minor data changes can lead to major structure changes; information gain in decision trees can be biased; and decision trees can often be relatively inaccurate. A popular alternative to decision trees is the influence diagram, which is a more compact, mathematical graphical representation of a decision situation.

Does HEAVY.AI Offer a Decision Tree Analysis Solution?

The HEAVY.AI platform natively integrates GPU-accelerated machine learning capabilities as part of its HeavyML module. Decision tree-models such as random forest, decision tree, and gradient boosted tree models, as well as other regression and clustering algorithms can be interactively trained, evaluated, and leveraged for predictive workflows, all directly from SQL and easily embeddable in Heavy Immerse.