Problem Statement

Apriori discovers frequent itemsets and association rules in transactional data. Used for market basket analysis. Currently missing from aprender.

Use Cases:

• Market basket analysis ("customers who bought X also bought Y")
• Recommendation systems
• Cross-selling strategies
• Web usage mining

Example Rules:

• {milk, bread} → {butter} (support=0.3, confidence=0.8)
• "30% of transactions contain milk, bread, butter"
• "80% of transactions with milk and bread also have butter"

Proposed Solution

Implement Apriori algorithm following EXTREME TDD.

Algorithm

Steps:

1. Find frequent 1-itemsets (items above min_support)
2. Generate candidate k-itemsets from frequent (k-1)-itemsets
3. Prune candidates using Apriori principle:
   • If itemset infrequent, all supersets are infrequent
4. Generate association rules from frequent itemsets
5. Filter rules by min_confidence

Implementation

API Design:

pub struct Apriori {
    min_support: f32,
    min_confidence: f32,
    frequent_itemsets: Option<Vec<ItemSet>>,
    rules: Option<Vec<AssociationRule>>,
}

pub struct ItemSet {
    items: Vec<usize>,
    support: f32,
}

pub struct AssociationRule {
    antecedent: Vec<usize>,  // If
    consequent: Vec<usize>,  // Then
    support: f32,
    confidence: f32,
    lift: f32,
}

impl Apriori {
    pub fn fit(&mut self, transactions: &[Vec<usize>]) -> Result<(), &'static str>;
    pub fn frequent_itemsets(&self) -> &[ItemSet];
    pub fn association_rules(&self) -> &[AssociationRule];
}

Success Criteria

• ✅ Apriori with frequent itemset mining
• ✅ Association rule generation
• ✅ Support, confidence, lift metrics
• ✅ 10+ tests (including retail dataset)
• ✅ Zero clippy warnings
• ✅ Example: examples/market_basket.rs

Estimated Effort

Timeline: 3-4 days
Complexity: Medium (combinatorial enumeration, pruning)