Association Rule Mining in Data Mining
What are Association Rules?
Association rules are simple “if–then” relationships found in data.
- If part (Antecedent) → condition
- Then part (Consequent) → result
Example:
“If a customer buys bread, then they also buy butter.”
These rules help us understand how items or events are related to each
other in large datasets such as:
- Shopping transactions
- Medical records
- Financial data
Use Cases of Association Rules
Association rules are widely used in many fields because they reveal hidden
patterns.
1. Market Basket Analysis
Used by retailers to find items that are often bought together
Example: Chips → Salsa
Helps in:
- Product placement
- Offers and discounts
2. Healthcare
Disease Diagnosis: Find patterns in symptoms and test results
Treatment Recommendation: Suggest better treatments based on patient
history
3. Financial Services
Fraud Detection: Identify unusual spending patterns
Cross-Selling: Recommend financial products to customers
4. Market Research
Analyze customer preferences
Improve advertising and product design
5. Web Usage Analysis
Understand user behavior on websites
Improve navigation and recommendations
6. Manufacturing
Detect causes of product defects
7. Telecommunications
Detect network issues
Predict customer churn (customers leaving service)
8. Inventory Management
Optimize stock levels Improve supply chain efficiency
9. Social Networks
Suggest friends or connections
10. Text Mining & Recommendation Systems
Recommend movies, books, or products (like Amazon, Netflix)
How Association Rules Work
The most common method used is the Apriori Algorithm.
Step 1:
Find Frequent Itemsets
Identify items that appear frequently together
Measured using Support
Example:
If milk appears in 40% of transactions → high support
Step 2:
Generate Rules
Create rules in the form:
If X → Then Y
Example:
If Milk → Then Bread
Step 3:
Rule Pruning (Filtering)
Keep only useful rules using:
Support: How often it appears
Confidence: How strong the rule is
Lift: Whether the relationship is meaningful
Step 4:
Repeat Process
The algorithm repeats steps to find better rules
Uses a key idea:
If a set is frequent, all its subsets are also frequent
Step 5:
Final Output A list of strong and useful rules
These rules help in decision-making
Measures to Evaluate Association Rules
1. Support
How often an itemset appears
Support(X)=Transactions containing X / Total Transactions
2. Confidence
Probability that Y occurs when X occurs
Confidence(X→Y)=Support(X) / Support(X∪Y)
3.Lift
Shows if X and Y are related or independent
Lift(X→Y)=Support(X∪Y) / Support(X)×Support(Y)
Lift > 1 → Positive relation
Lift < 1 → Negative relation
4. Interest (Correlation)
Difference between actual and expected occurrence
Interest(X→Y)=Support(X∪Y)−(Support(X)×Support(Y))
5. Conviction
Measures dependency of the rule
Conviction(X→Y)=1−Support(Y) / 1−Confidence(X→Y)
6. Leverage
Difference between observed and expected frequency
Leverage(X→Y)=Support(X∪Y)−(Support(X)×Support(Y))
Association Rule Algorithms
1. AIS Algorithm
Generates itemsets and counts them
Extends large itemsets step-by-step
2. SETM Algorithm
Stores transaction IDs
Generates itemsets after scanning database
3. Apriori Algorithm (Most Important)
Uses a bottom-up approach
Removes unnecessary itemset early
Based on rule:
“If a set is frequent, all its subsets are also frequent”
Uses of Association Rules
- Customer behavior analysis
- Market basket analysis
- Product recommendation
- Store layout design
- Fraud detection
- Machine learning models
Example of Association Rules
A famous example:
Diapers → Beer
Customers buying diapers are also likely to buy beer
Helps stores increase sales by smart placement
Application Areas
Association rules are used in:
- Retail
- Healthcare
- E-commerce
- Fraud Detection
- Web Analysis
- Inventory Management
- Text Mining
- Manufacturing
- Social Networks
- Telecommunications
- Customer Segmentation
Conclusion
Association Rule Mining helps discover hidden patterns and
relationships in data.
It plays a key role in:
- Better decision-making
- Personalized recommendations
- Business optimization
