R: Association Rules (arules) not writing any rules - is my dataset too small?

Question

I've read through Stack and various documentation online and I'm still not getting this to work.

I have a dataset of 5,368 transactions. They come in as an Excel sheet with a bunch of different columns - CustomerID, ItemID, and OrderID (see below, data comes in as it is shown from A1:C10).

I have 3 questions:

• Specifically what format does the data need to be in? I've tried reading it in using all 3 formats shown below. I can get read.transactions to read in the data in any of these formats, but when I go to run the apriori it just gives me 1 rule (or sometimes none). Even to get that one rule I have to set the confidence to .01 and the lhs is always blank.

• The most recent attempt I made, I used the format shown at row 21. I even cut out all the single transactions (row 23 & 24). I then ran this syntax:

sb<-read.transactions(file = "~/Downloads/sbasket.csv",sep = ",")

I think I even tried:

sb<-read.transactions(file = "~/Downloads/sbasket.csv", format="single",sep=",", cols=c(1,2))

• How big does your dataset need to be? My dataset is 5,368 rows, of which only 366 involve multiple transactions (so most lines in the data are like row 6 below), so roughly only 7% of my overall data has multiple item baskets. Is this why I am getting no rules? This is why I tried eliminating all single basket orders, but even that didn't help.

Answer 1

arules can read format 1 and 3. Use summary(sb) to check that the items are read-in correctly. Here is an example for your format 3:

trans_txt <- "13,19,20\n17\n1,\n16,17"
write(trans_txt, file = "trans.txt")

library("arules")
trans <- read.transactions("trans.txt", sep = ",")
summary(trans)

  transactions as itemMatrix in sparse format with
  4 rows (elements/itemsets/transactions) and
  6 columns (items) and a density of 0.2916667 

  most frequent items:
    17       1      13      16      19 (Other) 
     2       1       1       1       1       1 

  element (itemset/transaction) length distribution:
  sizes
  1 2 3 
  2 1 1 

  Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
  1.00    1.00    1.50    1.75    2.25    3.00 

  includes extended item information - examples:
  labels
  1      1
  2     13
  3     16

rules <- apriori(trans)
inspect(rules)

     lhs        rhs  support confidence lift count
[1]  {16}    => {17} 0.25    1          2    1    
[2]  {19}    => {20} 0.25    1          4    1    
[3]  {20}    => {19} 0.25    1          4    1    
[4]  {19}    => {13} 0.25    1          4    1    
[5]  {13}    => {19} 0.25    1          4    1    
[6]  {20}    => {13} 0.25    1          4    1    
[7]  {13}    => {20} 0.25    1          4    1    
[8]  {19,20} => {13} 0.25    1          4    1    
[9]  {13,19} => {20} 0.25    1          4    1    
[10] {13,20} => {19} 0.25    1          4    1 

Answer 2

I don't know anything about 'arules,' but is it possible the problem is it's expecting a csv and you are loading an excel spreadsheet? Maybe try using the package 'openxlsx' to read the file first, then input it to read.transactions?