Support & Confidence

INFO

Evaluates the strength of association rules based on the proportion of transactions that contain both antecedent and consequent items.

How It Works

For a rule $X\to Y$:

• Support:

  $$\text{Support}(X\to Y)=\frac{\text{Transactions containing }X\cup Y}{\text{Total transactions}}$$

• Confidence:

  $$\text{Confidence}(X\to Y)=\frac{\text{Transactions containing }X\cup Y}{\text{Transactions containing }X}$$

Support measures how frequently the rule occurs.
Confidence measures how reliably $Y$ occurs when $X$ does.

What to Look For

• High support = frequent pattern
• High confidence = strong conditional dependency
• Use thresholds to prune weak or rare rules

Application Models

• Apriori Algorithm
• Equivalence Class Clustering and Bottom-Up Lattice Traversal (ECLAT)