Question

I’m struggling to solve this problem in R.
I have data like this:

item   id
1      500
2      500
2      600
2      700
3      500
3      600

data.frame(item = c(1, 2, 2, 2, 3, 3),
           id = c(500, 500, 600, 700, 500, 600))

And I want to count the number of times a pair of items is linked to the same id.
So I want this output:

item1    item2    count
    1        2        1
    2        3        2
    1        3        2

I’ve tried approaching this with commands like:

x_agg = aggregate(x, by=list(x$id), c)

and then

x_agg_id = lapply(x_agg$item, unique)

thinking that I could then count the occurrence of each item. But the by function seems to create an object of lists, which I don’t know how to manipulate. I am hoping there is a simpler way….

Answer 1

# your data
df<-read.table(text="item   id
1      500
2      500
2      600
2      700
3      500
3      600",header=TRUE)


library(tnet)
item_item<-projecting_tm(df, method="sum")
names(item_item)<-c("item1","item2","count")

item_item

  #item1 item2 count
#1     1     2     1
#2     1     3     1
#3     2     1     1
#4     2     3     2
#5     3     1     1
#6     3     2     2

EDIT

how many ids and items do you have? you could always rename things. e.g.

numberitems<-length(unique(df$id))+9000
items<-data.frame(item=unique(df$item),newitems=c(9000:(numberitems-1)))
numberids<-length(unique(df$id))+1000
ids<-data.frame(id=unique(df$id),newids=c(1000:(numberids-1)))
newdf<-merge(df,items,by="item")
newdf<-merge(newdf,ids,by="id")
DF<-data.frame(item=newdf$newitems,id=newdf$newids)

library(tnet)
item_item<-projecting_tm(DF, method="sum")
names(item_item)<-c("item1","item2","count")

then merge back the original names afterwards….