Association Rule
In data mining, association rule mining is a popular and well-researched method for discovering interesting relationships between variables in a large database. Association rules are used to analyze and predict customer behavior but not restricted to healthcare, bioinformatics and etc. Association rule is if/then statement that helps to uncover relationships between unrelated data in the relational database or another information repository. It can be specified in linear implication expression like X⟹Y. where X and Y are itemsets.
Association Rule Example: Consider a customer who buys bread is likely to buy butter also and this statement can be expressed as.
Bread⟹Butter.
Such a statement can be used to express how items or objects are related to each other and how they tend to group together. Consider another example, if a customer buys a laptop and laptop sleeve then he likely to buy a wireless mouse. Such information can be used as a basis for marketing activities such as product promotion and product pricing.
Buys{Laptop, Laptop sleeve}⟹buys{Wireless mouse}.
A different section of association rule includes Antecedent (if), Consequent (then), Support and Confidence. From earlier example
Bread⟹Butter[10%,45%].
If a customer buys bread, then he likes to buy butter measured by percentage. In the above example, bread is antecedent, butter is consequent, 20% in support and 45% is confidence. Support and Confidence are two popular measurements used in association rule mining. Consider an association rule.
A⟹
B.
Support denotes the probability that contains both A and B. Confidence denotes the probability that a transaction containing A also contains B. For a better understanding of Confidence and Support consider another example. In the supermarket, the retailer wants to find the percentage of people who are buying bread by considering 100 total transactions. If 20 customer buys bread, then
.
Out of these 20 transactions, people who are buying bread also buy butter in 9 transactions, so confidence can be calculated as
.