The Bread Basket - Association Rule Mining
Divij Pherwani
Association Rule Mining is an unsupervised machine learning algorithm. The technique utilizes the apriori algorithm. The goal is to discover the association between the objects in datasets and common trends in the transactions.
The dataset used for analysis is “The Bread Basket” from Kaggle. The dataset belongs to a bakery located in Edinburgh. The dataset has 20507 entries, over 9000 transactions, and 4 columns. The dataset has transactions of customers who ordered different items from this bakery online and the time period of the data is from 26-01-11 to 27-12-03. Link to Kaggle dataset: (https://www.kaggle.com/mittalvasu95/the-bread-basket). There was no prior analysis done in R earlier and dataset meets the requirement of the project.
Loading Libraries and Data
# Loading all the libraries that will be used
library("plyr")
library(arules)## Loading required package: Matrix##
## Attaching package: 'arules'## The following objects are masked from 'package:base':
##
## abbreviate, writelibrary(arulesViz)## Loading required package: gridlibrary(tidyverse)## -- Attaching packages --------------------------------------- tidyverse 1.3.0 --## v ggplot2 3.3.3 v purrr 0.3.4
## v tibble 3.0.5 v dplyr 1.0.3
## v tidyr 1.1.2 v stringr 1.4.0
## v readr 1.4.0 v forcats 0.5.1## -- Conflicts ------------------------------------------ tidyverse_conflicts() --
## x dplyr::arrange() masks plyr::arrange()
## x purrr::compact() masks plyr::compact()
## x dplyr::count() masks plyr::count()
## x tidyr::expand() masks Matrix::expand()
## x dplyr::failwith() masks plyr::failwith()
## x dplyr::filter() masks stats::filter()
## x dplyr::id() masks plyr::id()
## x dplyr::lag() masks stats::lag()
## x dplyr::mutate() masks plyr::mutate()
## x tidyr::pack() masks Matrix::pack()
## x dplyr::recode() masks arules::recode()
## x dplyr::rename() masks plyr::rename()
## x dplyr::summarise() masks plyr::summarise()
## x dplyr::summarize() masks plyr::summarize()
## x tidyr::unpack() masks Matrix::unpack()library(lubridate)##
## Attaching package: 'lubridate'## The following objects are masked from 'package:arules':
##
## intersect, setdiff, union## The following objects are masked from 'package:base':
##
## date, intersect, setdiff, unionlibrary(dplyr)
library(plyr)
library(readxl)
library(xlsx)
library(lubridate)
library(data.table)##
## Attaching package: 'data.table'## The following objects are masked from 'package:lubridate':
##
## hour, isoweek, mday, minute, month, quarter, second, wday, week,
## yday, year## The following objects are masked from 'package:dplyr':
##
## between, first, last## The following object is masked from 'package:purrr':
##
## transposelibrary(splitstackshape)
setwd(dirname(rstudioapi::getSourceEditorContext()$path))The Transaction column contains the transaction ID of each transaction. The Item column contains the item name. The date_time column contains the timestamp of transaction. The period_day contains information about the period of day i.e., morning, afternoon, evening and night. The weekday_weekend column contains information if the day of the week is a weekday or weekend.
#Loading the dataset and displaying the header
df <- read.csv("bread basket.csv")
head(df)## Transaction Item date_time period_day weekday_weekend
## 1 1 Bread 30-10-2016 09:58 morning weekend
## 2 2 Scandinavian 30-10-2016 10:05 morning weekend
## 3 2 Scandinavian 30-10-2016 10:05 morning weekend
## 4 3 Hot chocolate 30-10-2016 10:07 morning weekend
## 5 3 Jam 30-10-2016 10:07 morning weekend
## 6 3 Cookies 30-10-2016 10:07 morning weekendExploratory Data Analysis
After loading the dataset, we can see that column Transaction is type numeric, column Item is type Character, column data_time is type Character, column period_day is type Character and column weekday_weekend is type Character.
# Summary of the bakery basket data frame
summary(df)## Transaction Item date_time period_day
## Min. : 1 Length:20507 Length:20507 Length:20507
## 1st Qu.:2552 Class :character Class :character Class :character
## Median :5137 Mode :character Mode :character Mode :character
## Mean :4976
## 3rd Qu.:7357
## Max. :9684
## weekday_weekend
## Length:20507
## Class :character
## Mode :character
##
##
## The dataset has 20507 entries and 5 columns.
# Another diagnostic for bakery basket data frame
str(df)## 'data.frame': 20507 obs. of 5 variables:
## $ Transaction : int 1 2 2 3 3 3 4 5 5 5 ...
## $ Item : chr "Bread" "Scandinavian" "Scandinavian" "Hot chocolate" ...
## $ date_time : chr "30-10-2016 09:58" "30-10-2016 10:05" "30-10-2016 10:05" "30-10-2016 10:07" ...
## $ period_day : chr "morning" "morning" "morning" "morning" ...
## $ weekday_weekend: chr "weekend" "weekend" "weekend" "weekend" ...The bakery sells 94 unique items to its customers.
# List of Unique Items sold by the bakery
unique(df$Item)## [1] "Bread" "Scandinavian"
## [3] "Hot chocolate" "Jam"
## [5] "Cookies" "Muffin"
## [7] "Coffee" "Pastry"
## [9] "Medialuna" "Tea"
## [11] "Tartine" "Basket"
## [13] "Mineral water" "Farm House"
## [15] "Fudge" "Juice"
## [17] "Ella's Kitchen Pouches" "Victorian Sponge"
## [19] "Frittata" "Hearty & Seasonal"
## [21] "Soup" "Pick and Mix Bowls"
## [23] "Smoothies" "Cake"
## [25] "Mighty Protein" "Chicken sand"
## [27] "Coke" "My-5 Fruit Shoot"
## [29] "Focaccia" "Sandwich"
## [31] "Alfajores" "Eggs"
## [33] "Brownie" "Dulce de Leche"
## [35] "Honey" "The BART"
## [37] "Granola" "Fairy Doors"
## [39] "Empanadas" "Keeping It Local"
## [41] "Art Tray" "Bowl Nic Pitt"
## [43] "Bread Pudding" "Adjustment"
## [45] "Truffles" "Chimichurri Oil"
## [47] "Bacon" "Spread"
## [49] "Kids biscuit" "Siblings"
## [51] "Caramel bites" "Jammie Dodgers"
## [53] "Tiffin" "Olum & polenta"
## [55] "Polenta" "The Nomad"
## [57] "Hack the stack" "Bakewell"
## [59] "Lemon and coconut" "Toast"
## [61] "Scone" "Crepes"
## [63] "Vegan mincepie" "Bare Popcorn"
## [65] "Muesli" "Crisps"
## [67] "Pintxos" "Gingerbread syrup"
## [69] "Panatone" "Brioche and salami"
## [71] "Afternoon with the baker" "Salad"
## [73] "Chicken Stew" "Spanish Brunch"
## [75] "Raspberry shortbread sandwich" "Extra Salami or Feta"
## [77] "Duck egg" "Baguette"
## [79] "Valentine's card" "Tshirt"
## [81] "Vegan Feast" "Postcard"
## [83] "Nomad bag" "Chocolates"
## [85] "Coffee granules " "Drinking chocolate spoons "
## [87] "Christmas common" "Argentina Night"
## [89] "Half slice Monster " "Gift voucher"
## [91] "Cherry me Dried fruit" "Mortimer"
## [93] "Raw bars" "Tacos/Fajita"This analysis is obvious as the day of the week is either a weekday or weekend.
# List of unique items in weekday_weekend colulmn
unique(df$weekday_weekend)## [1] "weekend" "weekday"This analysis is obvious as well, the periods of the day include morning, afternoon, evening and night.
# List of unique items in period_day column
unique(df$period_day)## [1] "morning" "afternoon" "evening" "night"There are no NA values in the dataset. While checking we iterating through each row and each column to find NA.
# Checking for Null/NA values in the dataset and there is none
df[rowSums(is.na(df)) > 0,]## [1] Transaction Item date_time period_day
## [5] weekday_weekend
## <0 rows> (or 0-length row.names)df[,colSums(is.na(df)) > 0]## data frame with 0 columns and 20507 rowsCoffee is by far the most popular item sold by the bakery. The other most popular items include Bread, Tea and Cake. Coffee culture in Edinburgh could be explored further via https://www.scotsman.com/lifestyle/food-and-drink/exploring-edinburghs-coffee-culture-1480578.
# Ten most popular items sold by the bakery
x <- as.data.frame(plyr::count(df, 'Item'))
x <- x %>% arrange(desc(freq))
x[1:10,]## Item freq
## 1 Coffee 5471
## 2 Bread 3325
## 3 Tea 1435
## 4 Cake 1025
## 5 Pastry 856
## 6 Sandwich 771
## 7 Medialuna 616
## 8 Hot chocolate 590
## 9 Cookies 540
## 10 Brownie 379Morning and afternoon is the most popular time for the transaction in bakery. The evening is not that popular and the transaction during the night at the bakery are almost non existent.
# Most popular period of day for bakery sale
y <- as.data.frame(plyr::count(df, 'period_day'))
y <- y %>% arrange(desc(freq))
y## period_day freq
## 1 afternoon 11569
## 2 morning 8404
## 3 evening 520
## 4 night 14The weekday has a higher frequency of transactions at the bakery than the weekend. It is not an apple to apple comparison. The frequency difference between weekend and weekday is not that significant.
# Frequency of weekday or weekend transaction
z <- as.data.frame(plyr::count(df, 'weekday_weekend'))
z <- z %>% arrange(desc(freq))
z## weekday_weekend freq
## 1 weekday 12807
## 2 weekend 7700The date_time column in broken into further columns to analyze the day more in depth. The column is split into date column, year column, month column and day column. Below we can see the head of the data frame after creating new columns.
# Breaking down date_time column in date column, time coulumn, year column, month column and day column
temp <- as.POSIXlt(df$date_time, format="%d-%m-%Y %H:%M")
df$year <- year(temp)
df$month <- month(temp)
df$date <- date(temp)
df$time <- as.ITime(temp, format = "%H:%M")
df$day <- weekdays(date(temp))
df$month <- month.abb[df$month]
head(df)## Transaction Item date_time period_day weekday_weekend year
## 1 1 Bread 30-10-2016 09:58 morning weekend 2016
## 2 2 Scandinavian 30-10-2016 10:05 morning weekend 2016
## 3 2 Scandinavian 30-10-2016 10:05 morning weekend 2016
## 4 3 Hot chocolate 30-10-2016 10:07 morning weekend 2016
## 5 3 Jam 30-10-2016 10:07 morning weekend 2016
## 6 3 Cookies 30-10-2016 10:07 morning weekend 2016
## month date time day
## 1 Oct 2016-10-30 09:58:00 Sunday
## 2 Oct 2016-10-30 10:05:00 Sunday
## 3 Oct 2016-10-30 10:05:00 Sunday
## 4 Oct 2016-10-30 10:07:00 Sunday
## 5 Oct 2016-10-30 10:07:00 Sunday
## 6 Oct 2016-10-30 10:07:00 SundayThe dataset contains bakery transactions from 30-10-2016 till 09-04-2017. There are almost 160 dates of transaction provided in the dataset.
# Dates analysis
unique(df$date)## [1] "2016-10-30" "2016-10-31" "2016-11-01" "2016-11-02" "2016-11-03"
## [6] "2016-11-04" "2016-11-05" "2016-11-06" "2016-11-07" "2016-11-08"
## [11] "2016-11-09" "2016-11-10" "2016-11-11" "2016-11-12" "2016-11-13"
## [16] "2016-11-14" "2016-11-15" "2016-11-16" "2016-11-17" "2016-11-18"
## [21] "2016-11-19" "2016-11-20" "2016-11-21" "2016-11-22" "2016-11-23"
## [26] "2016-11-24" "2016-11-25" "2016-11-26" "2016-11-27" "2016-11-28"
## [31] "2016-11-29" "2016-11-30" "2016-12-01" "2016-12-02" "2016-12-03"
## [36] "2016-12-04" "2016-12-05" "2016-12-06" "2016-12-07" "2016-12-08"
## [41] "2016-12-09" "2016-12-10" "2016-12-11" "2016-12-12" "2016-12-13"
## [46] "2016-12-14" "2016-12-15" "2016-12-16" "2016-12-17" "2016-12-18"
## [51] "2016-12-19" "2016-12-20" "2016-12-21" "2016-12-22" "2016-12-23"
## [56] "2016-12-24" "2016-12-27" "2016-12-28" "2016-12-29" "2016-12-30"
## [61] "2016-12-31" "2017-01-01" "2017-01-03" "2017-01-04" "2017-01-05"
## [66] "2017-01-06" "2017-01-07" "2017-01-08" "2017-01-09" "2017-01-10"
## [71] "2017-01-11" "2017-01-12" "2017-01-13" "2017-01-14" "2017-01-15"
## [76] "2017-01-16" "2017-01-17" "2017-01-18" "2017-01-19" "2017-01-20"
## [81] "2017-01-21" "2017-01-22" "2017-01-23" "2017-01-24" "2017-01-25"
## [86] "2017-01-26" "2017-01-27" "2017-01-28" "2017-01-29" "2017-01-30"
## [91] "2017-01-31" "2017-02-01" "2017-02-02" "2017-02-03" "2017-02-04"
## [96] "2017-02-05" "2017-02-06" "2017-02-07" "2017-02-08" "2017-02-09"
## [101] "2017-02-10" "2017-02-11" "2017-02-12" "2017-02-13" "2017-02-14"
## [106] "2017-02-15" "2017-02-16" "2017-02-17" "2017-02-18" "2017-02-19"
## [111] "2017-02-20" "2017-02-21" "2017-02-22" "2017-02-23" "2017-02-24"
## [116] "2017-02-25" "2017-02-26" "2017-02-27" "2017-02-28" "2017-03-01"
## [121] "2017-03-02" "2017-03-03" "2017-03-04" "2017-03-05" "2017-03-06"
## [126] "2017-03-07" "2017-03-08" "2017-03-09" "2017-03-10" "2017-03-11"
## [131] "2017-03-12" "2017-03-13" "2017-03-14" "2017-03-15" "2017-03-16"
## [136] "2017-03-17" "2017-03-18" "2017-03-19" "2017-03-20" "2017-03-21"
## [141] "2017-03-22" "2017-03-23" "2017-03-24" "2017-03-25" "2017-03-26"
## [146] "2017-03-27" "2017-03-28" "2017-03-29" "2017-03-30" "2017-03-31"
## [151] "2017-04-01" "2017-04-02" "2017-04-03" "2017-04-04" "2017-04-05"
## [156] "2017-04-06" "2017-04-07" "2017-04-08" "2017-04-09"min(unique(df$date))## [1] "2016-10-30"max(unique(df$date))## [1] "2017-04-09"The frequency of November is the highest, which would be month where bakery sold most items or there were most amount of transactions. October is the month with least transactions but it is understandable as the dataset begins from 30-10-2016. Most of the month was not documented.
# Frequency of transaction per month for bakery
x <- as.data.frame(plyr::count(df, 'month'))
x <- x %>% arrange(desc(freq))
x## month freq
## 1 Nov 4436
## 2 Mar 3944
## 3 Feb 3906
## 4 Jan 3356
## 5 Dec 3339
## 6 Apr 1157
## 7 Oct 3692017 is the year with highest frequency of transactions but it is an unfair comparison as the months documented in 2016 and 2017 are not equal. On the brighter side, bakery has stable transactions every month.
# Frequency of transaction per year for bakery
y <- as.data.frame(plyr::count(df, 'year'))
y <- y %>% arrange(desc(freq))
y## year freq
## 1 2017 12363
## 2 2016 8144Weekdays have the highest frequency, there are five days in weekdays and 2 in weekend so it is understandable as well. Saturday is the most popular day of the week with bakery transactions.
# Frequency of transaction per day of the week for bakery
z <- as.data.frame(plyr::count(df, 'day'))
z <- z %>% arrange(desc(freq))
z## day freq
## 1 Saturday 4605
## 2 Friday 3124
## 3 Sunday 3095
## 4 Thursday 2646
## 5 Tuesday 2392
## 6 Monday 2324
## 7 Wednesday 2321Data Engineering
After the analysis from the dataset, I am removing the unnecessary columns from the dataset and keeping only the items. The items are not in a single row, the same transaction id is present in multiple rows with item. So there is a need to document all the items for a transaction id in a single row to perform an easier analysis.
# create an empty data frame and feeding it with the filtered data from the earlier loaded data frame
colClasses = c("numeric", "character")
col.names = c("Transaction", "Items")
table <- read.table(text = "", colClasses = colClasses, col.names = col.names)
for (i in unique(df$Transaction))
{
x <- df[df$Transaction == i,2]
y <- ""
for (z in x)
{
z <- trimws(z)
z <- tolower(z)
if (y == "")
{
y <- paste0(y, z)
}
else
{
y <- paste(y, z, sep = " , ")
}
}
table[nrow(table)+1, ] <- list(i,y)
}
head(table)## Transaction Items
## 1 1 bread
## 2 2 scandinavian , scandinavian
## 3 3 hot chocolate , jam , cookies
## 4 4 muffin
## 5 5 coffee , pastry , bread
## 6 6 medialuna , pastry , muffinI remove the Transaction ID column and split the Items by “,” into new columns.
table <- cSplit(table, "Items", sep=",")
table <- table[,2:ncol(table)]
head(table)## Items_01 Items_02 Items_03 Items_04 Items_05 Items_06 Items_07
## 1: bread <NA> <NA> <NA> <NA> <NA> <NA>
## 2: scandinavian scandinavian <NA> <NA> <NA> <NA> <NA>
## 3: hot chocolate jam cookies <NA> <NA> <NA> <NA>
## 4: muffin <NA> <NA> <NA> <NA> <NA> <NA>
## 5: coffee pastry bread <NA> <NA> <NA> <NA>
## 6: medialuna pastry muffin <NA> <NA> <NA> <NA>
## Items_08 Items_09 Items_10 Items_11
## 1: <NA> <NA> <NA> <NA>
## 2: <NA> <NA> <NA> <NA>
## 3: <NA> <NA> <NA> <NA>
## 4: <NA> <NA> <NA> <NA>
## 5: <NA> <NA> <NA> <NA>
## 6: <NA> <NA> <NA> <NA>The engineered data is stored in a csv file which will be later analyzed for association rule mining in transaction.
# Writing to the new basket file
write.table(table, "basket.csv", col.names = FALSE, row.names=FALSE, na = "", sep = ",")Association Rule Mining
I begin by the reading the file as a transaction data and creating the transactional object.
bakery <- read.transactions("basket.csv", sep = ",")## Warning in asMethod(object): removing duplicated items in transactionssummary(bakery)## transactions as itemMatrix in sparse format with
## 9465 rows (elements/itemsets/transactions) and
## 94 columns (items) and a density of 0.02122827
##
## most frequent items:
## coffee bread tea cake pastry (Other)
## 4528 3097 1350 983 815 8114
##
## element (itemset/transaction) length distribution:
## sizes
## 1 2 3 4 5 6 7 8 9 10
## 3948 3059 1471 662 234 64 17 4 5 1
##
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.000 1.000 2.000 1.995 3.000 10.000
##
## includes extended item information - examples:
## labels
## 1 adjustment
## 2 afternoon with the baker
## 3 alfajoresBelow we can see the first ten elements of the sparse matrix.
inspect(bakery[1:10])## items
## [1] {bread}
## [2] {scandinavian}
## [3] {cookies,hot chocolate,jam}
## [4] {muffin}
## [5] {bread,coffee,pastry}
## [6] {medialuna,muffin,pastry}
## [7] {coffee,medialuna,pastry,tea}
## [8] {bread,pastry}
## [9] {bread,muffin}
## [10] {medialuna,scandinavian}Checking the support level of first 5 items in bakery data
itemFrequency(bakery[,1:5])## adjustment afternoon with the baker alfajores
## 0.0001056524 0.0045430534 0.0363444268
## argentina night art tray
## 0.0007395668 0.0040147913Frequency plot with set value of support at 10%
itemFrequencyPlot(bakery, support = 0.1)
Plot of top 15 items
itemFrequencyPlot(bakery, topN = 15)
Below is the visualization of Sparse Matrix for first 5 transactions.
image(bakery[1:5])
Randomly selecting 100 transaction samples for visualization of sparse matrix
image(sample(bakery, 100))
Using the apriori algorithm I find the association rules, the support is set at 1% and confidence is set at 25% with minimum length of rule being 2.
bakeryrules <- apriori(bakery, parameter = list(support = 0.01, confidence = 0.25, minlen = 2))## Apriori
##
## Parameter specification:
## confidence minval smax arem aval originalSupport maxtime support minlen
## 0.25 0.1 1 none FALSE TRUE 5 0.01 2
## maxlen target ext
## 10 rules TRUE
##
## Algorithmic control:
## filter tree heap memopt load sort verbose
## 0.1 TRUE TRUE FALSE TRUE 2 TRUE
##
## Absolute minimum support count: 94
##
## set item appearances ...[0 item(s)] done [0.00s].
## set transactions ...[94 item(s), 9465 transaction(s)] done [0.00s].
## sorting and recoding items ... [30 item(s)] done [0.00s].
## creating transaction tree ... done [0.00s].
## checking subsets of size 1 2 3 done [0.00s].
## writing ... [24 rule(s)] done [0.00s].
## creating S4 object ... done [0.00s].bakeryrules## set of 24 rulesBelow is the summary of the association rules.
summary(bakeryrules)## set of 24 rules
##
## rule length distribution (lhs + rhs):sizes
## 2 3
## 21 3
##
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 2.000 2.000 2.000 2.125 2.000 3.000
##
## summary of quality measures:
## support confidence coverage lift
## Min. :0.01004 Min. :0.2660 Min. :0.01817 Min. :0.5751
## 1st Qu.:0.01366 1st Qu.:0.3469 1st Qu.:0.03452 1st Qu.:0.8620
## Median :0.01965 Median :0.4899 Median :0.04004 Median :1.0304
## Mean :0.02640 Mean :0.4517 Mean :0.06397 Mean :1.0018
## 3rd Qu.:0.03098 3rd Qu.:0.5328 3rd Qu.:0.06432 3rd Qu.:1.1138
## Max. :0.09002 Max. :0.7044 Max. :0.32721 Max. :1.4724
## count
## Min. : 95.0
## 1st Qu.:129.2
## Median :186.0
## Mean :249.8
## 3rd Qu.:293.2
## Max. :852.0
##
## mining info:
## data ntransactions support confidence
## bakery 9465 0.01 0.25Inspecting first 10 bakery rules
inspect(bakeryrules[1:10])## lhs rhs support confidence coverage lift
## [1] {spanish brunch} => {coffee} 0.01088220 0.5988372 0.01817221 1.2517655
## [2] {toast} => {coffee} 0.02366614 0.7044025 0.03359746 1.4724315
## [3] {scone} => {coffee} 0.01806656 0.5229358 0.03454834 1.0931067
## [4] {soup} => {coffee} 0.01584786 0.4601227 0.03444268 0.9618068
## [5] {muffin} => {coffee} 0.01880613 0.4890110 0.03845747 1.0221928
## [6] {alfajores} => {bread} 0.01035394 0.2848837 0.03634443 0.8706569
## [7] {alfajores} => {coffee} 0.01965135 0.5406977 0.03634443 1.1302349
## [8] {brownie} => {bread} 0.01077655 0.2691293 0.04004226 0.8225085
## [9] {brownie} => {coffee} 0.01965135 0.4907652 0.04004226 1.0258596
## [10] {juice} => {coffee} 0.02060222 0.5342466 0.03856313 1.1167500
## count
## [1] 103
## [2] 224
## [3] 171
## [4] 150
## [5] 178
## [6] 98
## [7] 186
## [8] 102
## [9] 186
## [10] 195Inspecting first 5 bakery rules with decreasing order of lift
inspect(sort(bakeryrules, by = "lift")[1:5])## lhs rhs support confidence coverage lift
## [1] {toast} => {coffee} 0.02366614 0.7044025 0.03359746 1.472431
## [2] {spanish brunch} => {coffee} 0.01088220 0.5988372 0.01817221 1.251766
## [3] {medialuna} => {coffee} 0.03518225 0.5692308 0.06180666 1.189878
## [4] {pastry} => {coffee} 0.04754358 0.5521472 0.08610671 1.154168
## [5] {alfajores} => {coffee} 0.01965135 0.5406977 0.03634443 1.130235
## count
## [1] 224
## [2] 103
## [3] 333
## [4] 450
## [5] 186Inspecting first 5 bakery rules with decreasing order of confidence
inspect(sort(bakeryrules, by = "confidence")[1:5])## lhs rhs support confidence coverage lift
## [1] {toast} => {coffee} 0.02366614 0.7044025 0.03359746 1.472431
## [2] {spanish brunch} => {coffee} 0.01088220 0.5988372 0.01817221 1.251766
## [3] {medialuna} => {coffee} 0.03518225 0.5692308 0.06180666 1.189878
## [4] {pastry} => {coffee} 0.04754358 0.5521472 0.08610671 1.154168
## [5] {alfajores} => {coffee} 0.01965135 0.5406977 0.03634443 1.130235
## count
## [1] 224
## [2] 103
## [3] 333
## [4] 450
## [5] 186Inspecting first 5 bakery rules with decreasing order of support
inspect(sort(bakeryrules, by = "support")[1:5])## lhs rhs support confidence coverage lift count
## [1] {bread} => {coffee} 0.09001585 0.2751049 0.32720549 0.5750592 852
## [2] {cake} => {coffee} 0.05472795 0.5269583 0.10385631 1.1015151 518
## [3] {tea} => {coffee} 0.04986793 0.3496296 0.14263074 0.7308402 472
## [4] {pastry} => {coffee} 0.04754358 0.5521472 0.08610671 1.1541682 450
## [5] {sandwich} => {coffee} 0.03824617 0.5323529 0.07184363 1.1127916 362Inspecting first 5 bakery rules with decreasing order of count
inspect(sort(bakeryrules, by = "count")[1:5])## lhs rhs support confidence coverage lift count
## [1] {bread} => {coffee} 0.09001585 0.2751049 0.32720549 0.5750592 852
## [2] {cake} => {coffee} 0.05472795 0.5269583 0.10385631 1.1015151 518
## [3] {tea} => {coffee} 0.04986793 0.3496296 0.14263074 0.7308402 472
## [4] {pastry} => {coffee} 0.04754358 0.5521472 0.08610671 1.1541682 450
## [5] {sandwich} => {coffee} 0.03824617 0.5323529 0.07184363 1.1127916 362Preparing and inspecting rules for Coffee by setting Consequent as Coffee.
rules.coffee<-apriori(data=bakery, parameter=list(supp=0.01,conf = 0.25), appearance=list(default="lhs", rhs="coffee"), control=list(verbose=F))
rules.coffee.byconf<-sort(rules.coffee, by="confidence", decreasing=TRUE)
inspect(head(rules.coffee.byconf))## lhs rhs support confidence coverage lift
## [1] {toast} => {coffee} 0.02366614 0.7044025 0.03359746 1.472431
## [2] {spanish brunch} => {coffee} 0.01088220 0.5988372 0.01817221 1.251766
## [3] {medialuna} => {coffee} 0.03518225 0.5692308 0.06180666 1.189878
## [4] {pastry} => {coffee} 0.04754358 0.5521472 0.08610671 1.154168
## [5] {alfajores} => {coffee} 0.01965135 0.5406977 0.03634443 1.130235
## [6] {juice} => {coffee} 0.02060222 0.5342466 0.03856313 1.116750
## count
## [1] 224
## [2] 103
## [3] 333
## [4] 450
## [5] 186
## [6] 195Using another way to filter rules and inspecting all rules for Bread
breadrules <- subset(bakeryrules, items %in% "bread")
inspect(breadrules)## lhs rhs support confidence coverage lift count
## [1] {alfajores} => {bread} 0.01035394 0.2848837 0.03634443 0.8706569 98
## [2] {brownie} => {bread} 0.01077655 0.2691293 0.04004226 0.8225085 102
## [3] {cookies} => {bread} 0.01447438 0.2660194 0.05441099 0.8130041 137
## [4] {medialuna} => {bread} 0.01690438 0.2735043 0.06180666 0.8358792 160
## [5] {pastry} => {bread} 0.02916006 0.3386503 0.08610671 1.0349774 276
## [6] {bread} => {coffee} 0.09001585 0.2751049 0.32720549 0.5750592 852
## [7] {bread,pastry} => {coffee} 0.01119915 0.3840580 0.02916006 0.8028067 106
## [8] {bread,cake} => {coffee} 0.01003698 0.4298643 0.02334918 0.8985568 95I observed earlier during the EDA that coffee is the most popular item. Using the itemFrequencyPlot function I plot the item frequency. The plot below is absolute item frequency.
itemFrequencyPlot(bakery, topN=10, type="absolute", main="Item Frequency") 
I observed earlier during the EDA that coffee is the most popular item. Using the itemFrequencyPlot function I plot the item frequency. The plot below is relative item frequency.
itemFrequencyPlot(bakery, topN=10, type="relative", main="Item Frequency") 
Below is the scatter plot for all the 24 rules. On the x-axis, we have the support. On the y-axis, we have the confidence. The intensity of color in the plot signifies the lift value (darker the higher).
The relationship is not clear by the plot, there seems to be a correlation between confidence and lift.
plot(bakeryrules)
Below is the scatter plot for all rules for item coffee. On the x-axis, we have the support. On the y-axis, we have the confidence. The intensity of color in the plot signifies the lift value (darker the higher).
Higher the confidence, higher the lift.
plot(rules.coffee)
Below is the matrix plot for all bakery rules. On the x-axis, we have the Antecendent or LHS. On the y-axis, we have the Consequent or RHS. The intensity of color in the plot signifies the lift value (darker the higher).
Higher the Consequent, higher the lift. Higher the Antecedent, lower the lift.
plot(bakeryrules, method="matrix", measure="lift")## Itemsets in Antecedent (LHS)
## [1] "{toast}" "{spanish brunch}" "{juice}" "{sandwich}"
## [5] "{cake}" "{pastry}" "{scone}" "{hot chocolate}"
## [9] "{muffin}" "{medialuna}" "{alfajores}" "{soup}"
## [13] "{cookies}" "{brownie}" "{bread,cake}" "{cake,tea}"
## [17] "{bread,pastry}" "{tea}" "{bread}"
## Itemsets in Consequent (RHS)
## [1] "{bread}" "{coffee}"
The grouped plot is presented below for all rules. On the RHS, we have Consequent and on the LHS we have Antecedent. The size of the circles represent the support and the intensity of the color represent the lift.
plot(bakeryrules, method="grouped") 
The grouped plot is presented below for all coffee rules which are above 80% of the total. On the RHS, we have Consequent and on the LHS we have Antecedent. The size of the circles represent the support and the intensity of the color represent the lift.
plot(rules.coffee, method="grouped") 
I plot all the 24 rules for bakery for vizualization
plot(bakeryrules, method="graph")
I plot all the 20 rules for coffee in bakery for vizualization
plot(rules.coffee, method="graph")
Below we have Parallel coordinates plot for all the bakery rules. On the x-axis, we have position in the rule. On the y-axis, we have the nominal values. The support is represented by the width of the arrow line and the confidence is represented by the intensity of the color.
plot(bakeryrules, method="paracoord", control=list(reorder=TRUE))
Below we have Parallel coordinates plot for item coffee. On the x-axis, we have position in the rule. On the y-axis, we have the nominal values. The support is represented by the width of the arrow line and the confidence is represented by the intensity of the color.
plot(rules.coffee, method="paracoord", control=list(reorder=TRUE))
Below is an interactive chart to project the association rules network for relationship between rules and items.
plot(bakeryrules, method = "graph", measure = "lift", shading = "confidence", engine = "htmlwidget")Using eclat algorithm, we perform basic statistics with reference to confidence. The minimum support is set at 1 % with 5 items or less.
freq.items<-eclat(bakery, parameter=list(supp=0.01, maxlen=5)) ## Eclat
##
## parameter specification:
## tidLists support minlen maxlen target ext
## FALSE 0.01 1 5 frequent itemsets TRUE
##
## algorithmic control:
## sparse sort verbose
## 7 -2 TRUE
##
## Absolute minimum support count: 94
##
## create itemset ...
## set transactions ...[94 item(s), 9465 transaction(s)] done [0.00s].
## sorting and recoding items ... [30 item(s)] done [0.00s].
## creating sparse bit matrix ... [30 row(s), 9465 column(s)] done [0.00s].
## writing ... [61 set(s)] done [0.00s].
## Creating S4 object ... done [0.00s].inspect(freq.items)## items support transIdenticalToItemsets count
## [1] {coffee,spanish brunch} 0.01088220 103 103
## [2] {coffee,toast} 0.02366614 224 224
## [3] {coffee,scone} 0.01806656 171 171
## [4] {coffee,soup} 0.01584786 150 150
## [5] {coffee,muffin} 0.01880613 178 178
## [6] {alfajores,coffee} 0.01965135 186 186
## [7] {alfajores,bread} 0.01035394 98 98
## [8] {brownie,coffee} 0.01965135 186 186
## [9] {bread,brownie} 0.01077655 102 102
## [10] {coffee,juice} 0.02060222 195 195
## [11] {coffee,cookies} 0.02820919 267 267
## [12] {bread,cookies} 0.01447438 137 137
## [13] {coffee,medialuna} 0.03518225 333 333
## [14] {bread,medialuna} 0.01690438 160 160
## [15] {coffee,hot chocolate} 0.02958267 280 280
## [16] {bread,hot chocolate} 0.01341786 127 127
## [17] {cake,hot chocolate} 0.01141046 108 108
## [18] {coffee,sandwich} 0.03824617 362 362
## [19] {bread,sandwich} 0.01701004 161 161
## [20] {sandwich,tea} 0.01436873 136 136
## [21] {bread,coffee,pastry} 0.01119915 106 106
## [22] {coffee,pastry} 0.04754358 450 450
## [23] {bread,pastry} 0.02916006 276 276
## [24] {cake,coffee,tea} 0.01003698 95 95
## [25] {bread,cake,coffee} 0.01003698 95 95
## [26] {cake,coffee} 0.05472795 518 518
## [27] {bread,cake} 0.02334918 221 221
## [28] {cake,tea} 0.02377179 225 225
## [29] {coffee,tea} 0.04986793 472 472
## [30] {bread,tea} 0.02810354 266 266
## [31] {bread,coffee} 0.09001585 852 852
## [32] {coffee} 0.47839408 4528 4528
## [33] {bread} 0.32720549 3097 3097
## [34] {tea} 0.14263074 1350 1350
## [35] {cake} 0.10385631 983 983
## [36] {pastry} 0.08610671 815 815
## [37] {sandwich} 0.07184363 680 680
## [38] {hot chocolate} 0.05832013 552 552
## [39] {medialuna} 0.06180666 585 585
## [40] {cookies} 0.05441099 515 515
## [41] {juice} 0.03856313 365 365
## [42] {brownie} 0.04004226 379 379
## [43] {alfajores} 0.03634443 344 344
## [44] {muffin} 0.03845747 364 364
## [45] {soup} 0.03444268 326 326
## [46] {scone} 0.03454834 327 327
## [47] {toast} 0.03359746 318 318
## [48] {farm house} 0.03919704 371 371
## [49] {truffles} 0.02028526 192 192
## [50] {spanish brunch} 0.01817221 172 172
## [51] {scandinavian} 0.02905441 275 275
## [52] {coke} 0.01944004 184 184
## [53] {tiffin} 0.01542525 146 146
## [54] {mineral water} 0.01415742 134 134
## [55] {jammie dodgers} 0.01320655 125 125
## [56] {chicken stew} 0.01299525 123 123
## [57] {jam} 0.01500264 142 142
## [58] {salad} 0.01045959 99 99
## [59] {fudge} 0.01500264 142 142
## [60] {hearty & seasonal} 0.01056524 100 100
## [61] {baguette} 0.01605917 152 152Using eclat algorithm, we perform basic statistics with reference to confidence. The minimum support is set at 5 % with 5 items or less.
freq.items<-eclat(bakery, parameter=list(supp=0.05, maxlen=5)) ## Eclat
##
## parameter specification:
## tidLists support minlen maxlen target ext
## FALSE 0.05 1 5 frequent itemsets TRUE
##
## algorithmic control:
## sparse sort verbose
## 7 -2 TRUE
##
## Absolute minimum support count: 473
##
## create itemset ...
## set transactions ...[94 item(s), 9465 transaction(s)] done [0.00s].
## sorting and recoding items ... [9 item(s)] done [0.00s].
## creating bit matrix ... [9 row(s), 9465 column(s)] done [0.00s].
## writing ... [11 set(s)] done [0.00s].
## Creating S4 object ... done [0.00s].inspect(freq.items)## items support transIdenticalToItemsets count
## [1] {cake,coffee} 0.05472795 518 518
## [2] {bread,coffee} 0.09001585 852 852
## [3] {coffee} 0.47839408 4528 4528
## [4] {bread} 0.32720549 3097 3097
## [5] {tea} 0.14263074 1350 1350
## [6] {cake} 0.10385631 983 983
## [7] {pastry} 0.08610671 815 815
## [8] {sandwich} 0.07184363 680 680
## [9] {hot chocolate} 0.05832013 552 552
## [10] {medialuna} 0.06180666 585 585
## [11] {cookies} 0.05441099 515 515Using the S4 method, we create association rules with minimum confidence set at 10 %.
freq.rules<-ruleInduction(freq.items, bakery, confidence=0.1)
inspect(freq.rules)## lhs rhs support confidence lift itemset
## [1] {coffee} => {cake} 0.05472795 0.1143993 1.1015151 1
## [2] {cake} => {coffee} 0.05472795 0.5269583 1.1015151 1
## [3] {coffee} => {bread} 0.09001585 0.1881625 0.5750592 2
## [4] {bread} => {coffee} 0.09001585 0.2751049 0.5750592 2The rules are saved in a csv file.
# saving the output
write(bakeryrules, file = "bakeryrules.csv", sep = ",", quote = TRUE, row.names = FALSE)Summary
I used the apriori algorithm for association mining to mine rules for the bakery. Coffee is by far the most popular item at the bakery. It was hard to analysze this data because a lot of people just buy a single item at the bakery. People who buy toast are likely buy coffee with 70% confidence by far the highest and 1.4 lift value. People who buy bread buy coffee with the highest count. Bakery is for bread products but definitely coffee is one item the shop can’t get out of their menu. Also, plots were provided and along with complete analysis.
References:
Association Rule Mining in R, https://medium.com/swlh/association-rule-mining-in-r-acbd15e0de89
University of Warsaw, Unsupervised Learning Course by dr Jacek Lewkowicz