arules (version 1.7.15)

ruleInduction: Association Rule Induction from Itemsets

Description

Induces association rules that can be generated from supplied itemsets, optionally using a transactions data set to recount support.

Usage

ruleInduction(x, ...)
# S4 method for itemsets
ruleInduction(
  x,
  transactions = NULL,
  confidence = 0.8,
  method = c("ptree", "apriori"),
  reduce = FALSE,
  verbose = FALSE,
  ...
)

Value

A rules object containing all induced rules meeting the confidence threshold. Its quality data includes support, confidence, and lift; methods that recount transactions can include an itemset index identifying the source itemset.

Arguments

x

the set of itemsets from which rules will be induced.

...

unused; unknown arguments produce a warning.

transactions

the transactions used to mine x. This can be omitted for method = "ptree" only when x is a complete collection of frequent itemsets with support values.

confidence

numeric value in [0, 1] giving the minimum confidence threshold.

method

induction method: "ptree" or "apriori".

reduce

logical; remove unused items before counting to reduce memory use and potentially improve speed?

verbose

logical; report progress and timing information?

Author

Christian Buchta and Michael Hahsler

Details

All rules that can be created using the supplied itemsets and that surpass the specified minimum confidence threshold are returned. ruleInduction() can be used to produce closed association rules defined by Pei et al. (2000) as rules X => Y where both X and Y are closed frequent itemsets. See the code example in the Example section.

Rule induction implements two methods. The default is "ptree".

• "ptree" method without transactions: No transactions need to be specified if x contains a complete set of frequent itemsets. The itemsets' support counts are stored in a ptree and then retrieved to create rules and calculate confidence. This is very fast, but fails if support values are missing or x is not a complete set of frequent itemsets.

• "ptree" method with transactions: If transactions are specified then all transactions are counted into a prefix tree and later retrieved to create rules from the itemsets and calculate confidence values. This is slower, but necessary if x is not a complete set of frequent itemsets. To improve speed, unused items are removed from the transaction data before creating the prefix tree (this behavior can be changed using the argument reduce). This might be slower for large transaction data sets. However, this is highly recommended as the items are also reordered to reduce the counting time.

• "apriori" method (always needs transactions): All association rules are mined from the transactions data set using apriori() with the smallest support found in the itemsets. In a second step, all rules which cannot be generated from one of the itemsets are removed. This procedure is very slow, especially for itemsets with many elements or very low support.

References

Michael Hahsler, Christian Buchta, and Kurt Hornik. Selective association rule generation. Computational Statistics, 23(2):303-315, April 2008.

Jian Pei, Jiawei Han, Runying Mao. CLOSET: An Efficient Algorithm for Mining Frequent Closed Itemsets. ACM SIGMOD Workshop on Research Issues in Data Mining and Knowledge Discovery (DMKD 2000).

See Also

Other mining algorithms: APappearance-class, AScontrol-classes, ASparameter-classes, apriori(), eclat(), fim4r(), weclat()

Other postprocessing: is.closed(), is.generator(), is.maximal(), is.redundant(), is.significant(), is.superset()

Examples

data("Adult")

## find all closed frequent itemsets
closed_is <- apriori(Adult, target = "closed frequent itemsets", support = 0.4)
closed_is

## use rule induction to produce all closed association rules
closed_rules <- ruleInduction(closed_is, transactions = Adult, verbose = TRUE)

## inspect the resulting closed rules
summary(closed_rules)
inspect(head(closed_rules, by = "lift"))

## get rules from frequent itemsets. Here, transactions does not need to be
## specified for rule induction.
frequent_is <- eclat(Adult, support = 0.4)
assoc_rules <- ruleInduction(frequent_is)
assoc_rules
inspect(head(assoc_rules))

## for itemsets that are not a complete set of frequent itemsets,
## transactions need to be specified.
some_is <- sample(frequent_is, 10)
some_rules <- ruleInduction(some_is, transactions = Adult)
some_rules