Mining Association Rules in Large Databases , Apriori Algorithm,Association Rule Mining

1. Er. Nawaraj Bhandari
Data Warehouse/Data Mining
Mining Association Rules in Large
Databases
Chapter 7:

2. Introduction
Association rule mining finds interesting association or correlation relationships
among a large set of data items.
With massive amounts of data continuously being collected and stored , many
industries are becoming interested in mining association huge amounts of
business transaction records can help in many business decision making
processes, such as catalog design, cross-marketing, and loss-leader analysis.
 A typical example of association rule mining is market basket analysis.

3. Association Rules
 Analyzes and predicts customer behavior.
 If / then statements.
 Examples:
 Bread=>butter.
If someone purchase bread then he/she likely to purchase butter.
Buys{onions, potatoes}=> buys{tomatoes}

4. Parts of Association Rules
Bread=>butter[20%, 45%]
Bread: Antecedent
Butter: Consequent
20% is Support
And 45% is Confidence

5. Support and Confidence
A=>B
Support denoted probability that contains both A & B
Confidence denotes probability that a transaction
containing A also contains B.

6. Support and Confidence
Consider in a super market
Total transcations: 100
Bread: 20
So ,
20/100 * 100=20% which is support
In 20 transaction of bread, butter : 9 transactions
So, 9/20 * 100=45% which is confidence.

7. Types of Association Rules
Single dimension association rule
Multidimensional association rule
Hybrid association rule

8. Single dimension association rule
Bread=>Butter
Dimension: buying.
Here one and only dimension is buying.

9. Multi dimension association rule
 With 2 or more dimensions.
 Occupation(I.T), Age(>22)=>buys(laptops)
 Here we have 3 dimensions i.e occupation, age limit and buys.
 In multidimensional rules we can not duplicate dimension.

10. Hybrid dimension association rule
 Dimension or predicates can be repeated.
 Time(5 O'clock ), Buy(tea)=>Buy(biscuits)
 If a person at 5 o’clock get tea, he or she is likely to get biscuits also.
 Here dimensions are repeated.

11. Field of association rule
 Web usages mining
 Banking
 Bio informatics
 Market based analysis
 Credit/ debit card analysis
 Product clustering
 Catalog design

12. Algorithms of association rule
 Apriori Algorithm
 Elcat Algorithm
 F.P Growth Algorithm

13. Apriori Algorithm
 If you brought tooth brush, there will be suggestion of tooth paste or if you
brought beer there will be suggestion of chips and potato cracker etc.
 Many ecommerce websites are using these trends of suggestion in market. This
is called Apriori Algorithms. This is machine learning algorithms and a lot of
ecommerce websites (like flipcart, amazon) are using this.

14. Apriori Algorithm

15. Apriori Algorithm
Candidates First
C1:
Item Set Support Count
M 3
O 4
N 2
K 5
E 4
Y 3
D 1
A 1
U 1
C 2

16. Apriori Algorithm
L1: (The item set which are frequently repeating using minimum support)
Item Set Support Count
M 3
O 4
K 5
E 4
Y 3

17. Apriori Algorithm
Candidates First
C2:
Item Set Support Count
M, O 1
M, K 3
M, E 2
M,Y 2
O, K 3
O, E 3
O, Y 2
K, E 4
K, Y 3
E, Y 2

18. Apriori Algorithm
L2: (The item set which are frequently repeating using minimum support)
Item Set Support Count
M, K 3
O, K 3
O, E 3
K, E 4
K, Y 3

19. Apriori Algorithm
Candidates First
C3:
Item Set Support Count
M, K, O 1
M, K, E 2
M, K, Y 2
O, K, E 3
O, K, Y 2

20. Apriori Algorithm
L3: (The item set which are frequently repeating using minimum support)
Item Set Support Count
O, K, E 3

21. Apriori Algorithm
Now create association rules with support and confidence for O, K, E.
Association rules as like
O AND K GIVES E
Confidence= (support/no of time it occur i.e. O AND K OF O^K=>E)
For example confidence for o and k = (3/3)=1
Association Rule Support Confidence Confidence %
O^K=>E 3 3/3=1 100
O^E=>K 3 3/3=1 100
K^E=>O 3 3/4=0.75 75
E=>O^K 3 3/4=0.75 75
K=>O^E 3 3/5=0.6 60
O=>K^E 3 3/4=0.75 75

22. Apriori Algorithm
Compare this with the minimum confidence 80%
Association Rule Support Confidence Confidence %
O^K=>E 3 3/3=1 100
O^E=>K 3 3/3=1 100
Hence final association rules are:
O^K=>E
O^E=>K
Now this is called market basket analysis.

23. Pros and Cons of Association Rule Mining
Pros
 It is an easy-to-implement and easy-to-understand algorithm.
 It can be used on large itemsets.
Cons
 Sometimes, it may need to find a large number of candidate rules which can be
computationally expensive.
 Calculating support is also expensive because it has to go through the entire
database.
June 8, 2019 Data Mining: Concepts and Techniques 23

24. Assignment
Minimum support:2, Minimum confidence:70%. Use Apriori algorithm to get
frequent itemsets and strong association rules.
TID Item
1 I1, I3, I4
2 I2, I3, I5
3 I1, I2, I3, I5
4 I2, I5

25. References
1. Sam Anahory, Dennis Murray, “Data warehousing In the Real World”, Pearson
Education.
2. Kimball, R. “The Data Warehouse Toolkit”, Wiley, 1996.
3. Teorey, T. J., “Database Modeling and Design: The Entity-Relationship Approach”,
Morgan Kaufmann Publishers, Inc., 1990.
4. “An Overview of Data Warehousing and OLAP Technology”, S. Chaudhuri,
Microsoft Research
5. “Data Warehousing with Oracle”, M. A. Shahzad
6. “Data Mining Concepts and Techniques”, Morgan Kaufmann J. Han, M Kamber
Second Edition ISBN : 978-1-55860-901-3

26. ANY QUESTIONS?