Association Rule Mining

Use

Association rule mining discovers associations between independent items or events. That is, it determines sets of items that frequently appear togther in the dataset.

Example

Do customers who buy milk from a store tend to buy cereal at the same time?

Popular Techniques

The Apriori Algorithm and many techniques derived from it.

Measuring

The support and confidence measures of the learned rules. The apriori algorithm is also very simple and easy to manually trace.

Caveats

• The apriori algorithm makes finding association rules tractable, but it's still not very efficient. (Many more efficient variations have been proposed, but these complicate things).
• Some association rules may not make much sense, even if they're correct. In some cases, they may bring interesting but surprising hypotheses to light (beer and diapers were associated, for example).
• Association rule mining and clustering are two separate concepts, but can be confused at first glance.
• There's no process for factoring out chance relationships, though you can apply statistical significance tests to the results.