daltoolbox (version 1.3.787)

pat_dara: Divergent Association Ranking Analysis

Description

Rank attribute values whose ordering diverges between the dataset and discovered rules.

Usage

pat_dara(
  data,
  rules,
  rhs_attr,
  rhs_value = NULL,
  attributes = NULL,
  rule_weight = NULL
)

Value

A list of per-attribute data frames with columns value, dataset, rules, and c.

Arguments

data

Reference data frame.

rules

Tidy rule data frame produced by pat_dara_rules() or pat_rules_tidy().

rhs_attr

Right-hand-side attribute used by the mined rules.

rhs_value

Optional right-hand-side value used to restrict data.

attributes

Optional attributes to compare. Defaults to common columns in data and rules.

rule_weight

Optional numeric column in rules used as rule frequency instead of one count per rule.

Details

DARA compares two value-frequency orderings for each explanatory attribute: the ordering in a reference dataset, optionally restricted to a right-hand-side value, and the ordering in a table of association rules. The right-hand-side attribute is required so it can be excluded from explanatory divergence. The returned counter c increases for values that are more prominent in rules than in the reference data and decreases for values that are less prominent in rules.

Examples

if (requireNamespace("arules", quietly = TRUE)) {
  data("AdultUCI", package = "arules")
  data <- as.data.frame(AdultUCI)
  data <- data[, c("workclass", "education", "marital-status", "occupation", "income")]
  rules <- pat_dara_rules(data, rhs = "income=small", supp = 0.05, conf = 0.6)
  ranking <- pat_dara(data, rules, rhs_attr = "income", rhs_value = "small")
  ranking[["marital-status"]]
}