[R] 연관성 분석 (Association rule)

ar <- read.csv("AR.csv", header=T, encoding = 'UTF-8-BOM')

사용할 데이터를 불러옵니다.

ar$Personal_status___sex[ar$Personal_status___sex == '1'] <- 'divorced male'
ar$Personal_status___sex[ar$Personal_status___sex == '2'] <- 'divorced female'
ar$Personal_status___sex[ar$Personal_status___sex == '3'] <- 'single male'
ar$Personal_status___sex[ar$Personal_status___sex == '4'] <- 'married male'
ar$Personal_status___sex[ar$Personal_status___sex == '5'] <- 'single female'

ar$Age = ifelse(ar$Age < 20, 'age 10-20',
                ifelse(ar$Age < 30, 'age 20-30',
                       ifelse(ar$Age < 40, 'age 30-40',
                              ifelse(ar$Age < 50, 'age 40-50',
                                     ifelse(ar$Age < 60, 'age 50-60',
                                            ifelse(ar$Age < 70, 'age 60-70', 'age 70-80'))))))

ar$Personal_status___sex <- as.factor(ar$Personal_status___sex)
ar$Age <- as.factor(ar$Age)
ar$Credit_status <- as.factor(ar$Credit_status)

연속형 변수를 전부 범주형 변수로 바꿔줍니다.

library(arules)
rules <- apriori(ar[, c(2:4)], parameter = list(supp=0.2, conf=0.2))
inspect(rules)

arules 패키지에 있는 apriori 함수를 통해 규칙을 생성할 수 있습니다.

parameter를 통해 규칙에 제한조건을 줄 수 있습니다.

최소 support로 0.2, 최소 confidence로 0.2를 설정하였습니다.

rules <- subset(rules, subset = lift > 1) # lift > 1
rules.sorted <- sort(rules, by="lift")
inspect(rules.sorted)

     lhs                              rhs                           support confidence lift     count
[1]  {Age=[38,75]}                 => {Personal_status___sex=[3,4]} 0.245   0.7122093  1.112827 245  
[2]  {Personal_status___sex=[3,4]} => {Age=[38,75]}                 0.245   0.3828125  1.112827 245  
[3]  {Credit_status=N}             => {Age=[38,75]}                 0.256   0.3657143  1.063123 256  
[4]  {Age=[38,75]}                 => {Credit_status=N}             0.256   0.7441860  1.063123 256  
[5]  {Age=[28,38)}                 => {Personal_status___sex=[3,4]} 0.248   0.6794521  1.061644 248  
[6]  {Personal_status___sex=[3,4]} => {Age=[28,38)}                 0.248   0.3875000  1.061644 248  
[7]  {Personal_status___sex=[3,4]} => {Credit_status=N}             0.469   0.7328125  1.046875 469  
[8]  {Credit_status=N}             => {Personal_status___sex=[3,4]} 0.469   0.6700000  1.046875 469  
[9]  {Age=[28,38)}                 => {Credit_status=N}             0.260   0.7123288  1.017613 260  
[10] {Credit_status=N}             => {Age=[28,38)}                 0.260   0.3714286  1.017613 260  

subset 함수를 사용해서, lift > 1 조건을 만족하는 규칙들을 뽑아내었습니다.

그리고 sort 함수를 통해 lift가 높은 순으로 정렬하여 inspect 함수를 통해 출력하였습니다.