R - Affinity Analysis (Market Basket Analysis)

market_basket_analysis.r

#--------------------------------------------
# R affinity / market-basket analysis
#--------------------------------------------
# if someone picks X, how likely would they
# be to pick Y or YZ. Or, if they picked XY
# together, how likely they'd pick Z. Etc.
#
# This serves two fold purpose..
# 1) we can make a recommender system.. based on
#    things they picked together, or picked one and
#    it later led to a similar selection amongst population,
#    we can recommend that if a person chooses X, they
#    might also be interested in Y. EG: if they
#    bought cereal, they might also be highly interested
#    in buying milk. Or, (in the case of our data)
#    if they already had one class, they'd probably
#    be very interested in taking a follow up class
#    most folks seem interested in taking.
#
# 2) by seeing what groups together with high confidence,
#    we could entice customers to buy those things
#    together more by changing stores around to put
#    things with affinity near each other. This will
#    lead to more impulse buying of supportive items.

#--------------------------------------------
# Note to self 
#--------------------------------------------
# highlight word + F1
# in R-Studio to bring up help, then select
# help for info on usage of keywords / funcs.
#
# Also, need to run R-Studio as admin for
# it to be able to install pkgs, or else
# get constant pop-up asking for personal
# lib dir that doesn't work.

#--------------------------------------------
# Setup / Load Lib(s)
#--------------------------------------------
# lib installed through R-Studio.
# already installed, so skipping,
# but leaving code in case needing
# to update / refresh lib later.

#install.packages("arules")

# load lib for use

library(arules)


#--------------------------------------------
#     Pre-processing
#--------------------------------------------

# normally we'd do a bit of cleanup & prep,
# but data set is small enough to just scan
# and make sure no nan's and odd formatting.
# ... skipping ...


#--------------------------------------------
# 1.	Load data into R.
#--------------------------------------------

# debug... where are we?
#getwd()

# R uses fwd slashes for dir structure in names, so...
# wrong
#filename = "D:\R Projects\Affinity\student-courses.csv"
# right
filename = "D:/R Projects/Affinity/student-courses.csv"

# load data
#dataload = read.csv ( filename )

# DEBUG.. check contents
#dataload


# OOPS .. we're loading data
# in an alternate way as transactionals
# so... comment out above part, and...

# read in as "single" format
#datatrans <- read.transactions ( filename, format = "basket", sep = ",")
datatrans <- read.transactions ( filename, format = "single", sep = ",", cols = c(1,2) )

# DEBUG... double-check trans loadout
#inspect ( datatrans )

##############################
# student data is formatted as a traditional table
# (student_id, course_id).
#
# so, we need "single" format, b/c it will convert
# the traditional data into a market basket format
# during load so apriori can use it below
#
# EG:
#
# so... "basket" format with data as-is gives us ...
#
#[768]  {319,AC102}        
#[769]  {319,AC113}        
#[770]  {319,IT741} 
#
# which won't work with apriori below
#
# instead, we use "single" to have it basket-ize data the
# traditional data...
#
#[8]    {AC111,CO810,FI201,MG501,QA812}                   1004         
#[9]    {MB113,MB114,MB202}                               1005         
#[10]   {EC403}                                           1006  
#
# .. which is what apriori wants


# from https://www.datacamp.com/community/tutorials/market-basket-analysis-r
# DEBUG... using summary() to get overview of data
#summary(datatrans)


#--------------------------------------------
#     Post-processing
#--------------------------------------------

# again.. data set is small enough, and with
# what we're doing with it there's no need to
# post-proc anything to prep for analysis below.
# ... skipping ...


#--------------------------------------------
# 2.	Perform affinity analysis using:
#     - support of 0.02
#     - confidence of 0.3
#--------------------------------------------

rules <- apriori( datatrans, parameter = list ( supp = 0.02, conf = 0.3 ) )



################################
# OUTPUT (from "basket" format)
# Oops ... NO RULES
# (this was when I realized "basket" format wouldn't work
# and had to figure things out)
################################
# Parameter specification:
#   confidence minval smax arem  aval originalSupport maxtime support minlen maxlen
# 0.3    0.1    1 none FALSE            TRUE       5     0.2      1     10
# target   ext
# rules FALSE
# 
# Algorithmic control:
#   filter tree heap memopt load sort verbose
# 0.1 TRUE TRUE  FALSE TRUE    2    TRUE
# 
# Absolute minimum support count: 849 
# 
# set item appearances ...[0 item(s)] done [0.00s].
# set transactions ...[1800 item(s), 4248 transaction(s)] done [0.00s].
# sorting and recoding items ... [0 item(s)] done [0.00s].
# creating transaction tree ... done [0.00s].
# checking subsets of size 1 done [0.00s].
# writing ... [0 rule(s)] done [0.00s].
# creating S4 object  ... done [0.00s].
################################

################################
# OUTPUT (from "single" format)
#... RULES! .. Ta-DA!
################################
# Parameter specification:
#   confidence minval smax arem  aval originalSupport maxtime support minlen maxlen target   ext
# 0.3    0.1    1 none FALSE            TRUE       5    0.02      1     10  rules FALSE
# 
# Algorithmic control:
#   filter tree heap memopt load sort verbose
# 0.1 TRUE TRUE  FALSE TRUE    2    TRUE
# 
# Absolute minimum support count: 32 
# 
# set item appearances ...[0 item(s)] done [0.00s].
# set transactions ...[154 item(s), 1646 transaction(s)] done [0.00s].
# sorting and recoding items ... [35 item(s)] done [0.00s].
# creating transaction tree ... done [0.00s].
# checking subsets of size 1 2 3 done [0.00s].
# writing ... [40 rule(s)] done [0.00s].
# creating S4 object  ... done [0.00s].
################################




#--------------------------------------------
# 3.	How many rules did the system generate? ... Answ = 40
#--------------------------------------------

inspect(rules)

################################
# ID    lhs             rhs     support    confidence lift      count
# [1]  {AC113}       => {AC102} 0.02247874 0.9736842  41.094467 37   
# [2]  {AC102}       => {AC113} 0.02247874 0.9487179  41.094467 37   
# [3]  {AC111}       => {AC115} 0.02187120 0.8181818  22.077496 36   
# [4]  {AC115}       => {AC111} 0.02187120 0.5901639  22.077496 36   
# [5]  {FI211}       => {FI227} 0.02308627 0.6785714  18.930993 38   
# [6]  {FI227}       => {FI211} 0.02308627 0.6440678  18.930993 38   
# [7]  {FI211}       => {MG503} 0.02065614 0.6071429   6.406136 34   
# [8]  {MG501}       => {EC408} 0.02247874 0.3557692   4.839638 37   
# [9]  {EC408}       => {MG501} 0.02247874 0.3057851   4.839638 37   
# [10] {MG501}       => {QA812} 0.02187120 0.3461538   4.632270 36   
# [11] {MG501}       => {MK601} 0.03219927 0.5096154   5.745390 53   
# [12] {MK601}       => {MG501} 0.03219927 0.3630137   5.745390 53   
# [13] {MG501}       => {FI201} 0.03159174 0.5000000   5.920863 52   
# [14] {FI201}       => {MG501} 0.03159174 0.3741007   5.920863 52   
# [15] {MG501}       => {AC104} 0.03159174 0.5000000   5.143750 52   
# [16] {AC104}       => {MG501} 0.03159174 0.3250000   5.143750 52   
# [17] {CO810}       => {MG503} 0.03098420 0.5257732   5.547581 51   
# [18] {MG503}       => {CO810} 0.03098420 0.3269231   5.547581 51   
# [19] {EC408}       => {MK601} 0.02673147 0.3636364   4.099626 44   
# [20] {MK601}       => {EC408} 0.02673147 0.3013699   4.099626 44   
# [21] {EC408}       => {FI201} 0.02673147 0.3636364   4.306082 44   
# [22] {FI201}       => {EC408} 0.02673147 0.3165468   4.306082 44   
# [23] {EC408}       => {AC104} 0.02794654 0.3801653   3.910950 46   
# [24] {QA812}       => {MK601} 0.02551640 0.3414634   3.849649 42   
# [25] {QA812}       => {FI201} 0.03948967 0.5284553   6.257823 65   
# [26] {FI201}       => {QA812} 0.03948967 0.4676259   6.257823 65   
# [27] {QA812}       => {AC104} 0.04131227 0.5528455   5.687398 68   
# [28] {AC104}       => {QA812} 0.04131227 0.4250000   5.687398 68   
# [29] {MK601}       => {FI201} 0.03645200 0.4109589   4.866463 60   
# [30] {FI201}       => {MK601} 0.03645200 0.4316547   4.866463 60   
# [31] {MK601}       => {AC104} 0.03827461 0.4315068   4.439127 63   
# [32] {AC104}       => {MK601} 0.03827461 0.3937500   4.439127 63   
# [33] {FI201}       => {AC104} 0.03766707 0.4460432   4.588669 62   
# [34] {AC104}       => {FI201} 0.03766707 0.3875000   4.588669 62   
# [35] {FI201,QA812} => {AC104} 0.02308627 0.5846154   6.014231 38   
# [36] {AC104,QA812} => {FI201} 0.02308627 0.5588235   6.617435 38   
# [37] {AC104,FI201} => {QA812} 0.02308627 0.6129032   8.201941 38   
# [38] {FI201,MK601} => {AC104} 0.02065614 0.5666667   5.829583 34   
# [39] {AC104,MK601} => {FI201} 0.02065614 0.5396825   6.390773 34   
# [40] {AC104,FI201} => {MK601} 0.02065614 0.5483871   6.182501 34 
################################

rules

# set of 40 rules


#--------------------------------------------
# 4.	Sort the display by descending order of
#     lift, and display the top 5 rules.
#--------------------------------------------

rules <- sort ( rules, decreasing = TRUE, by = "lift" )
inspect(rules[0:5])

################################
# lhs             rhs    support    confidence lift     count
# [1] {AC113} => {AC102} 0.02247874 0.9736842  41.09447 37   
# [2] {AC102} => {AC113} 0.02247874 0.9487179  41.09447 37   
# [3] {AC111} => {AC115} 0.02187120 0.8181818  22.07750 36   
# [4] {AC115} => {AC111} 0.02187120 0.5901639  22.07750 36   
# [5] {FI211} => {FI227} 0.02308627 0.6785714  18.93099 38 
################################



#--------------------------------------------
# 5.	If a student has taken AC104, which course
#     (just one) would you recommend to this
#     student?
#
#     If you had to recommend two (2), which
#      would they be? 
#--------------------------------------------

inspect(rules)

# couldn't figure out how to filter for AC104,
# so pruned out a list manually from lift sort list

# ID    lhs             rhs     support    confidence lift      count
# [7]  {AC104,FI201} => {QA812} 0.02308627 0.6129032   8.201941 38   
# [8]  {AC104,QA812} => {FI201} 0.02308627 0.5588235   6.617435 38   
# [10] {AC104,MK601} => {FI201} 0.02065614 0.5396825   6.390773 34   
# [13] {AC104,FI201} => {MK601} 0.02065614 0.5483871   6.182501 34   
# [21] {AC104}       => {QA812} 0.04131227 0.4250000   5.687398 68   
# [24] {AC104}       => {MG501} 0.03159174 0.3250000   5.143750 52   
# [31] {AC104}       => {FI201} 0.03766707 0.3875000   4.588669 62   
# [34] {AC104}       => {MK601} 0.03827461 0.3937500   4.439127 63   

#----------
# 1 class recommendation
#----------

# [7]  {AC104,FI201} => {QA812} 0.02308627 0.6129032   8.201941 38 

# QA812

#----------
# 2 class recommendations
#----------

# [7]  {AC104,FI201} => {QA812} 0.02308627 0.6129032   8.201941 38   
# [8]  {AC104,QA812} => {FI201} 0.02308627 0.5588235   6.617435 38 

# QA812 & FI201

student-courses.csv