Transforming a table of Nearest Neighbor Distances into a Matrix

Question

I have a data frame generated from computing nearest neighbor (K=2) using the RANN package. I would like to transform this data into a matrix with values of 0,1,2 for each cell with 0 = not neighbor, 1=nearest neighbor, 2=2nd nearest neighbor.

The data frame has two columns, the first column is the ID of the 1st NN, the second column is the ID of the 2nd NN. The rows correspond to the ID of the point from which the NN were calculated.

Is there an existing routine to easily to this sort of transformation?

Thanks

[Reproducibility](http://stackoverflow.com/a/5963610/1317221) is great. — user1317221_G, Jan 26 '13 at 20:57

user1317221_G · Accepted Answer · 2013-01-26T21:29:04.337

2

Based on the limited idea you have given, here is, I think, an unpretty solution:

 NNdf <- data.frame(NN1=c(1,2,4),NN2=c(2,3,1)) # make up some data
 NNdf$origin <- rownames(NNdf)

 NNdf
#  NN1 NN2 origin
#1   1   2      1
#2   2   3      2
#3   4   1      3


 library(reshape2)
 hold <- melt(NNdf, id = "origin")
 hold
#  origin variable value
#1      1      NN1     1
#2      2      NN1     2
#3      3      NN1     4
#4      1      NN2     2
#5      2      NN2     3
#6      3      NN2     1


 hold2 <- dcast(hold, origin~value, value.var="variable")

 hold2[hold2 == "NN1"]  <- 1
 hold2[hold2 == "NN2"]  <- 2
 hold2[is.na(hold2) ]   <- 0

 hold2
#  origin 1 2 3 4
#1      1 1 2 0 0
#2      2 0 1 2 0
#3      3 2 0 0 1

(this might rely on apply(hold2,1,as.numeric) afterwards)

edited Jan 26 '13 at 21:29

answered Jan 26 '13 at 21:20

user1317221_G

15,087
3
52
78

This worked perfectly, if not a bit heavy on the processing. Thanks and sorry for not including sample data. – scuerda Jan 27 '13 at 14:11
glad it worked, just to say it might be better to turn the `NN1` etc to numbers before the `dcast` rather than after. So `hold <- within(hold, variable <- ifelse(variable == "NN1",1,ifelse(variable == "NN2",2,0)))` before you make `hold2` – user1317221_G Jan 27 '13 at 14:17

score 0 · Answer 2 · answered Jan 26 '13 at 21:57

Another possibility, but not especially prettier. Thanks to @user1317221_G for the sample data !

NNdf <- data.frame(NN1=c(1,2,4,3),NN2=c(2,3,1,2))
NNdf$origin <- as.numeric(rownames(NNdf))
NNdf

  NN1 NN2 origin
1   1   2      1
2   2   3      2
3   4   1      3
4   3   2      4

res <- matrix(0,nrow(NNdf),nrow(NNdf))
res[as.matrix(NNdf[,c("origin","NN1")])] <- 1
res[as.matrix(NNdf[,c("origin","NN2")])] <- 2
res

     [,1] [,2] [,3] [,4]
[1,]    1    2    0    0
[2,]    0    1    2    0
[3,]    2    0    0    1
[4,]    0    2    1    0

Transforming a table of Nearest Neighbor Distances into a Matrix

2 Answers2