Association

TL;DR

• A relationship between two or more entities that can be discovered and represented in data.
• Common representations include correlation measures and association rules.
• Metrics like support, confidence, and lift quantify how often and how strongly items appear together and can be used to predict future behavior.

Definition

Association is a type of relationship that exists between two or more entities. This relationship can be represented using various techniques in data science, such as correlation or the construction of association rules.

Explanation

Association captures how entities (for example, items in transactions) relate to one another through co-occurrence or statistical dependence. It can be quantified with metrics such as support, which measures the frequency with which a set of items appear together in a dataset. Association can also be expressed as association rules, which state relationships in if-then form and can be evaluated with metrics like confidence and lift.

Examples

Grocery store example

• A customer who frequently buys bread and milk is likely to also buy butter. In this case, the items “bread”, “milk”, and “butter” are associated because they are often purchased together.
• The support for the set of items {bread, milk, butter} is the number of transactions in which all three items were purchased.
• An association rule for this dataset might be: “IF a customer buys bread and milk, THEN they are likely to also buy butter”. This rule can be used to predict that a customer who buys bread and milk is likely to also buy butter.

Use cases

• Using association rules and related metrics to make predictions about future behavior (for example, predicting items a customer is likely to buy based on their current purchases).

Notes or pitfalls

• Support measures how frequently a set of items appears together in the dataset.
• Confidence measures the likelihood that an association rule will hold true.
• Lift measures the degree to which the presence of one item affects the likelihood of the other items in the set.

Related terms

• Correlation
• Association rules
• Support
• Confidence
• Lift