In [2]:
library(arules)
library(dplyr)
In [41]:
par(family ="NanumBarunGothic")
In [3]:
tr <- read.delim("dataTransactions.tab", stringsAsFactors=FALSE)
head(tr)
- 일반식품, 화장품 제외
- 가장 많이 팔린 Data
- 돌려봐야 계속 나오게 된다.
- Custid와 Corner 가 중복이 되지 않도록 추출.
In [5]:
tr.filter <- tr %>%
filter(!(corner %in% c("일반식품","화장품"))) %>%
distinct(custid, corner)
In [6]:
head(tr.filter)
- Custid별로 corner 데이터를 짤라 List로 형성.
In [7]:
head(split(tr.filter$corner, tr.filter$custid))
In [19]:
# custid별로 corner를 자른다.
trans <- as(split(tr.filter$corner, tr.filter$custid), "transactions") #transactions 메소드
trans
In [8]:
# trans <- read.transactions("dataTransactions.tab", format = "single", sep="\t", cols = c(2,6), skip=1)
# 일반식품, 화장품 제거없이 하려면 위와 같이 해도 상관 없다.
Example of Transactions from a Matrix¶
In [9]:
a_matrix <- matrix(
c(1,1,1,0,0,
1,1,0,0,0,
1,1,0,1,0,
0,0,1,0,1,
1,1,0,1,1), ncol = 5)
In [10]:
dimnames(a_matrix) <- list(
c("a","b","c","d","e"),
paste("Tr",c(1:5), sep = ""))
a_matrix
- Transactions : count of rows
- 사용자의 수라고 보면 될 것같다. ( 위의 데이터에 )
- Items : count of columns
- 아이템 수
In [11]:
trans2 <- as(a_matrix, "transactions")
trans2
- 각 사용자가 포함한 Item을 list형태로 ( R의 list는 아님 ) 만들어 놓음. 집합 형태로
In [12]:
inspect(trans2)
Example of transaction from data.frame¶
In [13]:
a_df <- data.frame(
age = as.factor(c(6,8,7,6,9,5)),
grade = as.factor(c(1,3,1,1,4,1)))
In [14]:
a_df
In [15]:
trans3 <- as(a_df, "transactions")
inspect(trans3)
In [16]:
options(repr.plot.width=4,repr.plot.height=3)
image(trans3)
Continue Lectures¶
In [20]:
inspect(trans[1:2])
- size(trans) each transaction length
In [21]:
transactionInfo(trans[size(trans) > 20])
In [22]:
image(trans[1:5])
In [23]:
options(repr.plot.width=4,repr.plot.height=5)
image(sample(trans, 100, replace = FALSE), main = "matrix diagram") # Sampling
- Table()과 동일한 결과를 출력한다.
In [24]:
t(itemFrequency(trans, type="absolute"))
In [25]:
table(tr.filter$corner)
In [26]:
t(round(itemFrequency(trans)[order(itemFrequency(trans), decreasing = TRUE)],2))
In [43]:
options(repr.plot.width=4,repr.plot.height=4)
itemFrequencyPlot(trans, support=0.2, cex.names=0.8,family = "HYsanB")
In [44]:
itemFrequencyPlot(trans, topN = 20, main = "support top 20 items")
- rules <- apriori(trans, parameter=list(support=0.2, confidence=0.8), appearance=list(rhs="스포츠",default="lhs"))
- 결과에 스포츠만 만든다.
In [27]:
rules <- apriori(trans, parameter=list(support=0.2, confidence=0.8))
summary(rules)
| rule1 | rule2 | rule3 |
|---|---|---|
| 2 | 3 | 4 |
| 1 | 40 | 29 |
| a->b | a+b->c | a+b+c->d |
In [28]:
inspect(rules)
- 뻔한 얘기들이 많이 나온다.
- 즉, Level Down이 필요하다.
In [29]:
inspect(sort(rules, by = "lift")[1:30])
In [30]:
rules.target <- subset(rules, rhs %in% "스포츠" & lift > 1.4)
In [31]:
inspect(sort(rules.target, by="confidence"))
In [32]:
rule.interest <- subset(rules, items %in% c("장신구", "섬유"))
inspect(rule.interest[1:10])
In [32]:
write(rules.target, file="arules.csv", sep=",", row.name=F)
In [35]:
library(pmml) # 세계 표준 문서
In [36]:
write.PMML(rules.target, file = "arules.xml")
Visualize Association Rules using arulesViz package¶
In [38]:
library(arulesViz)
In [39]:
plot(rules)
In [43]:
plot(sort(rules, by = "lift")[1:20], method = "grouped")
In [44]:
plot(rules, method = "graph", control = list(type="items"))
Exercise¶
In [3]:
data <- read.delim("shoppingmall.txt", stringsAsFactors=FALSE)
In [4]:
head(data,3)
In [6]:
st <- as.matrix(data[,-1])
st[1:5,]
In [7]:
trans <- as(st, "transactions")
In [46]:
inspect(trans[1:20])
In [49]:
options(repr.plot.width=4,repr.plot.height=3)
image(trans[1:5])
In [47]:
inspect(trans[1:2])
In [51]:
options(repr.plot.width=4,repr.plot.height=7)
image(sample(trans, 100, replace = FALSE), main = "matrix diagram")
In [52]:
rules <- apriori(trans, parameter=list(support=0.01, confidence=0.8))
In [53]:
summary(rules)
In [57]:
inspect(sort(rules, by = "lift")[1:20])