Association Rules/Market Basket Brainstorming.

[vudo41]

Hi! I wanted to understand more about arules and association rules/MarketBasket.

What is this data telling me exactly? What is a good range of confidence range to use?

How can I make my reporting more understandable?

#instal packages needed to perform Market Basket Analysis
#Data was downloaded from DART with timeframe 1/1/2019 - 4/30/2019

install.packages(c("arules", "arulesViz", "tidyverse", "readxl", "plyr", "ggplot2", "knitr", "lubridate", "ddply"))

#read excel ito R dataframe
library(readxl)
Customer_Retail_Data <- read_excel("C:/Users/avu/Desktop/Marketing Directives/Customer Retail Data 2019 YTD 4302019.xlsx")

#view Data
View(Customer_Retail_Data)

#Use ddply to group OrderID and Date, and then to be in a singles format (combine all products from that CustomerID and Date as one row, with each item, seperated by a ',')
library(plyr)
transactionData <- ddply(Customer_Retail_Data,c("CustomerID","Date"),
function(df1)paste(df1$CategoryID,
collapse = ","))

#view Data
View(transactionData)

#as OrderID and Date will not be of any use in rule mining, so I am setting them to NULL
#set column OrderID of dataframe TransactionData
transactionData$CustomerID <- NULL

#set column Date of dataframe TransactionData
transactionData$Date <-NULL

#rename column to items

colnames(transactionData) <- c("items")

#show dataframe transactionData. This format is called "basket" format.
View(transactionData)

#transactionData: Data to be written
write.csv(transactionData,"C:/Users/avu/Desktop/Marketing Directives/CustomerRetailData2019YTD.csv", quote = FALSE, row.names = FALSE)

#Load this transaction data into an object of transaction class
library(arules)
ctr <- read.transactions('C:/Users/avu/Desktop/Marketing Directives/CustomerRetailData2019YTD.csv', format = 'basket', sep=',')

#data is right skewed
summary(ctr)

Create an item frequency plot for the top 20 items

if (!require("RColorBrewer")) {

install color package of R

install.packages("RColorBrewer")
#include library RColorBrewer
library(RColorBrewer)
}
itemFrequencyPlot(ctr,topN=20,type="absolute",col=brewer.pal(8,'Pastel2'), main="Absolute Item Frequency Plot")

itemFrequencyPlot(ctr,topN=20,type="relative",col=brewer.pal(8,'Pastel2'),main="Relative Item Frequency Plot")

Min Support as 0.001, confidence as 0.8.

association.rules <- apriori(ctr, parameter = list(supp=0.001, conf=0.6,minlen=1))

summary(association.rules)
inspect(association.rules[1:285])

[EconomiCurtis]

@vudo41

("C:/Users/avu/Desktop/Marketing Directives/Customer Retail Data 2019 YTD 4302019.xlsx")

No one here is going to have access to this file, making it hard to dig into your problem. You'll improve your chances of finding good help here if you ask this question with a shiny-reprex (details: Shiny debugging and reprex guide), ideally linking to a rstudio.cloud project.

[andresrcs]

When you inspect association rules you get three metrics, confidence, support and lift. confidence is an indication of how often the rule has been found to be true, support is an indication of how frequently the itemset appears in the dataset. and lift lets us know the degree to which those two occurrences are dependent on one another, and makes those rules potentially useful for predicting the consequent in future data sets.

@vudo41

What is a good range of confidence range to use?

I think this is going to depend on your particular application and domain specific knowledge, so there is no general answer for this.