Ming-Chang Lee

Topic : Data mining – R - association rules and apriori algorithm

Author : Ming-Chang Lee
Date : 2009.03.30

# Topic : Data mining - association rules and apriori algorithm

# Author : Ming-Chang Lee
# Date : 2009.03.30
# Note : 資料來源, 參考 northwind.mdb 北風資料庫

## step 1. load arules package

library(arules)

## step 2. Import and transform data

workpath <- "C:/temp"

setwd(workpath) # Set working directory. Make c:\temp directory and
copy csv file to the directory
getwd() # Get working directory

# import raw data (csv file)

nw_data <- read.table(file="northwind_trans.csv", header = TRUE, sep
= "," )
class(nw_data) # "data.frame"
mode(nw_data) # "list"

# group data by orderID. using column 1(OrderID) and 2(ProductName) only

nw_temp <- tapply(nw_data[,2], nw_data[,1], function(x) paste(x))
nw_temp[1] # get the first row data
names(nw_temp[1]) # get the name of first row data
nw_temp[[1]] # get the element of first row data
class(nw_temp) # "array"
mode(nw_temp) # "list"

nw2 <- vector("list", length(nw_temp)) # length(nw_temp)=830

for (i in seq(nw_temp)) names(nw2)[i] <- names(nw_temp[i])
for (i in seq(nw_temp)) nw2[[i]] <- nw_temp[[i]]
class(nw2) # "list"
mode(nw2) # "list"

1
 Ming-Chang Lee

# force data into transactions

nw <- as(nw2, "transactions")
class(nw) # "transactions"
mode(nw) # "S4"

# step 3. analyze data

# generate level plots to visually inspect binary incidence matrices
image(nw) # result - Figure 1 Level plot
summary(nw)

# step 4. find 1-items (L1)

# provides the generic function itemFrequency and the frequency/support
for all single items in an objects based on itemMatrix.
itemFrequency(nw, type = "relative") # default: "relative"
itemFrequency(nw, type = "absolute")

# step 5.
# create an item frequency bar plot for inspecting the item frequency
distribution for objects based on itemMatrix
itemFrequencyPlot(nw) # result- Figure 2 Item frequency bar plot

# step 6.
# mine association rules
# rules <- apriori(nw) # Mine association rules using default Apriori
algorithm
rules1 <- apriori(nw, parameter = list(supp = 0.05, target = "maximally
frequent itemsets")) # set parameters
rules2 <- apriori(nw, parameter = list(supp = 0.001, conf = 0.8, target
= "rules")) # set parameters

# step7.
# display results
inspect(nw) # display transactions
inspect(rules1) # display maximally frequent itemset
inspect(rules2[1]) # display association

# reference:

2
 Ming-Chang Lee

# Data Mining: Han, J. and Kamber, M. (2006) Concepts and Techniques,

Second Edition, Morgan Kaufmann.
# http://r-forge.r-project.org/projects/arules
# end

分析結果:
> inspect(rules1) # display maximally frequent itemset
items support
1 {item062} 0.05783133
2 {item056} 0.06024096
3 {item075} 0.05542169
4 {item059} 0.06506024
5 {item041} 0.05662651
6 {item016} 0.05180723
7 {item031} 0.06144578
8 {item024} 0.06265060
9 {item002} 0.05421687
10 {item060} 0.06144578
> inspect(rules2[1]) # display association
lhs rhs support confidence lift
1 {item048,
item061} => {item043} 0.001204819 1 29.64286
>

item048, item061 => item043 {support=0.12%, confidence=1}

Note: 僅列出第一個 rule

3
 item frequency (relative)

2.
1.
0.00 0.01 0.02 0.03 0.04 0.05 0.06

ite
item
item001
item0002
item003
item 0

Reference
item0045
item0006
item007
item008
item 09
item010
item0011
item012
item013
item 14
item015
item0016
item017
item018
item 19
item020
item0021
item022
item 2
item0234
item025
item0026

Edition, Morgan Kaufmann.

item027
item 28
item029
item0030
item031
item032
item 33
item034 Row
item0035
item 36
800
600
400
200

item037

4
item0038
item039
item040
20 60

Column
item 41
item042
item0043
Dimensions: 830 x 77

item044
item045
item 46
item047 Figure 1 Level plot
item0048
item049
item050
item 51
item052
item0053
item054

Figure 2 Item frequency bar plot

item 5
item0556
item0057
item058
item059
item 60
item061

R – arules Package, http://r-forge.r-project.org/projects/arules

item062
item0063
item064
item 65
item066
item0067
item068
item069
item 70
item071
item0072
item073
item 7
item00745
m 76
07
7

Data Mining: Han, J. and Kamber, M. (2006) Concepts and Techniques, Second
Ming-Chang Lee