93 R Recommendation Algorithms - Association Rules

1 Initial preparation

Prepare the environment

 clean up
rm(list=ls())

# load package arules to find rules
library(arules)

2 Read the file

Read the file, check the content of the file, and find that the three tables that need to be given are: transaction information table, commodity list, and commodity order table.
According to the PA field, the transaction information table and the commodity order table are connected together.

read purchase data on items level

trans <- read.csv('purchase.csv',header = TRUE,stringsAsFactors = FALSE )
RA_desc <- read.csv('RA_desc.csv',header = TRUE,stringsAsFactors = FALSE )
PA_desc <- read.csv('PA_desc.csv',header = TRUE,stringsAsFactors = FALSE )
dim(trans)
head(trans)
head(PA_desc)
names(trans) <- c("card","transaction_place","transaction_time","pos_id","transaction_number","invo_num","amt_aft","amt_bft","quantity","PA","RA","product_code")

# merge with descriptions

trans2 <- merge(trans,PA_desc, by="PA", all.x=TRUE)

#write trans_is and items

trans2$trans_id = paste(trans2$card, trans2$transaction_place, trans2$transaction_time,
                        trans2$pos_id, trans2$transaction_number, sep='_')
View(head(trans2))



write.csv(subset(trans2,select=c("trans_id","PA_desc")), file = 'purchase2.csv', row.names = F)

trans3 = read.transactions('purchase2.csv', format = "single", sep = ",",
                           cols = c("trans_id", "PA_desc"), rm.duplicates=TRUE,header = TRUE)

Finally, organize the information into order number and shopping item details

trans3 = read.transactions('purchase2.csv', format = "single", sep = ",",
cols = c("trans_id", "PA_desc"), rm.duplicates=TRUE, header = TRUE)
This step is Convert data into transaction order format for easy algorithm processing

3 Apriori algorithm generation rules

find rules with low confidence level

rules.all = apriori(trans3,
                   parameter = list(minlen=2,
                                    supp=0.008,
                                    conf=0.5,
                                    target = "rules"))
inspect(rules.all)



# find rules with reasonable confidence level
rules = apriori(trans3,
               parameter = list(minlen=2,
                                supp=0.005,
                                conf=0.05,
                                target = "rules"))
inspect(rules)

write(rules,"rules.csv",row.names=FALSE,sep="|")

The previous association algorithm rule has dealt with possible redundant rules, so the rules are pruned

adjust the format of results

quality(rules.all) <- round(quality(rules.all), digits=3)
rules.sorted <- sort(rules.all, by ='lift')
inspect(rules.sorted)

reules.pruned = rules.sorted
inspect(reules.pruned[1])

target:
a character string indicating the type of association mined. One of
“frequent itemsets”
“maximally frequent itemsets”
“closed frequent itemsets”
“rules” (only available for Apriori; use ruleInduction for eclat.)
“hyper edge sets” (only available for Apriori; see references for the definition of association hyperedgesets)

4 Association rule drawing

scatter plots

# plot rules.all
png(file="rules_all.png", bg="white", width=1200, height=900)
plot(rules.all)
dev.off()

Drawing of various rules

plot(reules.pruned)
plot(reules.pruned, measure=c("support", "lift"), shading="confidence")

plot(reules.pruned, shading="order", control=list(main = "Two-key plot"))

inspect(reules.pruned)

other drawings

# sel <- plot(rules, measure=c("support", "lift"), shading="confidence", interactive=TRUE)

### matrix plots
plot(reules.pruned, method="matrix", measure="lift")
plot(reules.pruned, method="matrix", measure="lift", control=list(reorder=TRUE))

plot(reules.pruned, method="matrix", measure=c("lift", "confidence"))
plot(reules.pruned, method="matrix", measure=c("lift", "confidence"),control=list(reorder=TRUE))


### grouped matrix plots
# sel <- plot(reules.pruned, method="white", interactive=TRUE)
png(file="grouped_matrix.png", bg="transparent", width=1200, height=900)
plot(reules.pruned, method='grouped')
dev.off()

plot(reules.pruned, method='grouped',control=list(k=25))

### graph plots
subrules2 <- head(sort(reules.pruned, by="lift"), 25)

inspect(subrules2)
plot(subrules2, method="graph")
plot(subrules2, method="graph", control=list(type="items"))
plot(subrules2, method='graph', control = list(type="itemsets"))

parallel coordinates

plot(subrules2, method=‘paracoord’, control = list(reorder=T))

subrules3 <- head(sort(reules.pruned, by=“lift”), 3)
plot(subrules3, method=‘paracoord’, control = list(reorder=T))
inspect(subrules3)