Association Rule Mining In Machine Learning
1) Introduction
Association
Rule Mining is a data-mining technique used to discover relationships
between items in large datasets. It answers questions like:
- “If a customer buys A,
what else do they usually buy?”
- “Which services are commonly
used together?”
- “What combinations of events
happen frequently?”
It is
most famous in Market Basket Analysis (retail), but it is used in many
other domains too.
2) Key Idea (in simple words)
Association
rules look like this:
A → B
Meaning:
If A happens, then B also tends to happen (in the same
transaction/session/event).
Example:
- {Bread, Butter} → {Jam}
Customers who buy bread and butter often also buy jam.
3) When Association Rules are Used
Use
association rule mining when:
✅ Your data is transactional (each row is a basket / set / group
of items)
✅ You want to find co-occurrence patterns
✅ Order of items does not matter (this is not sequence mining)
✅ Your goal is recommendations, cross-sell, bundling, or pattern
discovery
Typical
datasets:
- Shopping baskets (items
bought together)
- Website sessions (pages
visited in the same session)
- Medical prescriptions (drugs
taken together)
- Bank transactions (services
used together)
- Course registrations
(subjects chosen together)
4) Real-Life Examples Where It’s Required
Example 1: Retail (Classic Market Basket)
A
supermarket wants to increase sales by bundling products.
- Rule found: {Milk} → {Bread}
- Action: Place milk and bread
closer, or show bread offer to milk buyers.
Example 2: E-Commerce Recommendations
Amazon/Flipkart-style
recommendations:
- Rule found: {Mobile Phone} → {Screen
Guard, Back Cover}
- Action: “Frequently bought
together” suggestions.
Example 3: Healthcare / Pharmacy
Identify
frequently co-prescribed medicines:
- Rule: {Diabetes Medicine} → {Blood
Pressure Medicine}
- Action: Better stock
planning, clinical insight checks.
Example 4: Education / Course Selection
University
finds which courses students choose together:
- Rule: {Python} → {Data Science}
- Action: Recommend learning
pathways.
Example 5: Banking / Financial Services
Find
which services are commonly used by same customer:
- Rule: {Savings Account} → {Debit
Card, Mobile Banking}
- Action: targeted product
offers.
5) Core Terms You Must Know
(A) Support
How
frequently an
itemset appears in the dataset.
Example:
If 100 transactions exist and 20 contain {Milk}, then
Support(Milk) = 20/100 = 0.20
(B) Confidence
How often
B appears when A appears.
Confidence(A
→ B) = Support(A ∪ B) / Support(A)
Example:
If Milk appears in 20 baskets, and Milk+Bread appears in 12 baskets,
Confidence(Milk → Bread) = 12/20 = 0.60
(C) Lift
Shows
whether A and B occur together more than chance.
Lift(A →
B) = Confidence(A → B) / Support(B)
Interpretation:
- Lift > 1: Positive association
(useful rule)
- Lift = 1: No real association
- Lift < 1: Negative association
0 टिप्पण्या
कृपया तुमच्या प्रियजनांना लेख शेअर करा आणि तुमचा अभिप्राय जरूर नोंदवा. 🙏 🙏