Association rules associate a particular conclusion (the purchase of a particular product, for example) with a set of conditions (the purchase of several other products, for example).

For example, the rule

beer <= cannedveg & frozenmeal (173, 17.0%, 0.84)

states that beer often occurs when cannedveg and frozenmeal occur together. The rule is 84% reliable and applies to 17% of the data, or 173 records. Association rule algorithms automatically find the associations that you could find manually using visualization techniques, such as the Web node.

The advantage of association rule algorithms over the more standard decision tree algorithms (C5.0 and C&R Trees) is that associations can exist between any of the attributes. A decision tree algorithm will build rules with only a single conclusion, whereas association algorithms attempt to find many rules, each of which may have a different conclusion.

The disadvantage of association algorithms is that they are trying to find patterns within a potentially very large search space and, hence, can require much more time to run than a decision tree algorithm. The algorithms use a generate and test method for finding rules--simple rules are generated initially, and these are validated against the dataset. The good rules are stored and all rules, subject to various constraints, are then specialized. Specialization is the process of adding conditions to a rule. These new rules are then validated against the data, and the process iteratively stores the best or most interesting rules found. The user usually supplies some limit to the possible number of antecedents to allow in a rule, and various techniques based on information theory or efficient indexing schemes are used to reduce the potentially large search space.

At the end of the processing, a table of the best rules is presented. Unlike a decision tree, this set of association rules cannot be used directly to make predictions in the way that a standard model (such as a decision tree or a neural network) can. This is due to the many different possible conclusions for the rules. Another level of transformation is required to transform the association rules into a classification rule set. Hence, the association rules produced by association algorithms are known as unrefined models. Although the user can browse these unrefined models, they cannot be used explicitly as classification models unless the user tells the system to generate a classification model from the unrefined model. This is done from the browser through a Generate menu option.

Two association rule algorithms are supported:

• The Apriori node extracts a set of rules from the data, pulling out the rules with the highest information content. Apriori offers five different methods of selecting rules and uses a sophisticated indexing scheme to process large data sets efficiently. For large problems, Apriori is generally faster to train; it has no arbitrary limit on the number of rules that can be retained, and it can handle rules with up to 32 preconditions. Apriori requires that input and output fields all be categorical but delivers better performance because it is optimized for this type of data.
• The Sequence node discovers association rules in sequential or time-oriented data. A sequence is a list of item sets that tends to occur in a predictable order. For example, a customer who purchases a razor and aftershave lotion may purchase shaving cream the next time he shops. The Sequence node is based on the CARMA association rules algorithm, which uses an efficient two-pass method for finding sequences.