Logical Interpretation
A decision tree can also be viewed as a logical expression, not just a geometric model.
1. Key Idea
A decision tree represents:
👉 OR of AND conditions
- Each path from root → leaf = AND (conjunction)
- The whole tree = OR (disjunction) of all such paths
2. What is a Conjunction (AND)?
A conjunction means:
Condition1 AND Condition2 AND Condition3
Example:
Outlook = Sunny AND Humidity = High
👉 This corresponds to one path in the tree
3. What is a Disjunction (OR)?
A disjunction means:
Path1 OR Path2 OR Path3
👉 This combines multiple paths
4. Example Tree Logic
Imagine a simple decision tree:
IF Outlook = Sunny AND Humidity = High → No
IF Outlook = Rain AND Wind = Weak → Yes
This can be written as:
\[
(Outlook = \text{Sunny} \land Humidity = \text{High})
\lor
(Outlook = \text{Rain} \land Wind = \text{Weak})
\]
👉 This is called Disjunctive Normal Form (DNF)
5. Why this matters
- Helps convert trees → rules
- Makes models interpretable
- Useful in expert systems & rule engines
Logical Form
\[
(Outlook = Sunny \land Humidity = High)
\lor
(Outlook = Rain \land Wind = Weak)
\]
5. General Mathematical Form
For a tree with k leaf nodes:
\[
f(x) = \bigvee_{i=1}^{k} \left( \bigwedge_{j=1}^{m_i} condition_{ij} \right)
\]
Where:
- ( i ) = path index
- ( j ) = condition index in a path
- ( m_i ) = number of conditions in path ( i )
6. Interpretation
- Each conjunction defines a region of the input space
- The disjunction combines all regions that lead to the same prediction
7. Advantages of Logical Representation
- Easy to interpret
- Can be converted into if-then rules
- Useful for rule-based systems
- Helps in explaining model decisions
8. Connection to Geometry
- Logical AND → Intersection of regions
- Logical OR → Union of regions
9. Summary
| Concept | Meaning |
|---|---|
| Path | Conjunction (AND) |
| Tree | Disjunction (OR) |
| Representation | Disjunctive Normal Form |
10. Conclusion
A decision tree can be viewed as a set of logical rules:
A collection of AND conditions combined using OR
This makes decision trees one of the most interpretable machine learning models.