Association Rule Learning is a rule-based machine learning method for discovering interesting relationships between variables in large datasets. It's a form of unsupervised learning that identifies patterns in the form of 'if-then' rules, such as 'If a customer buys {X}, they are also likely to buy {Y}.' The strength of these rules is evaluated using metrics like support, confidence, and lift, which measure the frequency, reliability, and strength of the association, respectively.
The concept of association rule learning was first introduced in a 1993 paper by Rakesh Agrawal, Tomasz Imieliński, and Arun Swami. However, it was the development of the Apriori algorithm in 1994 that popularized the technique and made it practical for use in large-scale data mining tasks, particularly for 'market basket analysis.'
Association rule learning is the core of many recommendation engines, most famously in e-commerce for features like 'Customers who bought this also bought...'. It's also used in a variety of other fields, including medical diagnosis (to identify relationships between symptoms and diseases), web usage mining (to analyze user behavior), and bioinformatics (to find patterns in genetic data).