arules

arules

Mining association rules and frequent itemsets.

Transactions

library(arules)

# From list
trans_list <- list(
  c("milk", "bread", "butter"),
  c("milk", "bread"),
  c("milk", "eggs"),
  c("bread", "butter", "eggs")
)
trans <- as(trans_list, "transactions")

# From data frame (binary)
trans <- as(df, "transactions")

# From CSV
trans <- read.transactions("basket.csv", format = "basket", sep = ",")
trans <- read.transactions("single.csv", format = "single", cols = c(1, 2))

Inspect Transactions

# Summary
summary(trans)

# View items
inspect(trans[1:5])

# Item frequency
itemFrequency(trans)
itemFrequency(trans, type = "absolute")

# Plot frequency
itemFrequencyPlot(trans, topN = 20)

Apriori Algorithm

# Mine frequent itemsets
itemsets <- apriori(trans,
  parameter = list(
    support = 0.01,
    target = "frequent itemsets"
  )
)

# Mine association rules
rules <- apriori(trans,
  parameter = list(
    support = 0.01,
    confidence = 0.5,
    minlen = 2
  )
)

Rule Parameters

rules <- apriori(trans,
  parameter = list(
    support = 0.01,      # Minimum support
    confidence = 0.5,    # Minimum confidence
    minlen = 2,          # Minimum items in rule
    maxlen = 10,         # Maximum items in rule
    target = "rules"     # "rules", "frequent itemsets", "maximally frequent itemsets"
  )
)

Inspect Rules

# Summary
summary(rules)

# View rules
inspect(rules)
inspect(head(sort(rules, by = "lift"), 10))

# Quality measures
quality(rules)

Rule Measures

# Support: P(A ∪ B)
# Confidence: P(B|A) = P(A ∪ B) / P(A)
# Lift: P(B|A) / P(B)

# Additional measures
interestMeasure(rules, c("chiSquared", "conviction", "leverage"), trans)

Filtering Rules

# By quality
high_conf <- subset(rules, confidence > 0.8)
high_lift <- subset(rules, lift > 2)

# By items
milk_rules <- subset(rules, items %in% "milk")
lhs_milk <- subset(rules, lhs %in% "milk")
rhs_milk <- subset(rules, rhs %in% "milk")

# Redundant rules
non_redundant <- rules[!is.redundant(rules)]