CODEWITHSUNDEEP
r
cs0815
cs0815

Reputation: 17418

using nearest_neighbor_search of distance package

I am trying to use the distances package with code along those lines:

library(distances)
library(dplyr)

set.seed(42)

x <- matrix(rnorm(100), ncol = 10)

df <- as.data.frame(x)
df$id <- paste0("x_", seq.int(nrow(df)))

distances <- distances(df, id_variable = "id")
distances

This produces a distances matrix object like this:

          x_1      x_2      x_3      x_4      x_5      x_6      x_7      x_8      x_9     x_10
x_1  0.000000 3.843183 4.093911 3.643060 4.935399 4.327867 4.287775 6.205355 6.197274 2.180997
x_2  3.843183 0.000000 5.084690 5.170822 5.067423 3.788407 4.383991 5.770031 7.113060 2.830447
x_3  4.093911 5.084690 0.000000 3.571286 4.547878 4.102882 3.531970 3.916854 6.470266 3.733713
x_4  3.643060 5.170822 3.571286 0.000000 3.820931 3.842954 3.667172 5.513104 5.176213 3.294032
x_5  4.935399 5.067423 4.547878 3.820931 0.000000 4.815130 3.465038 5.917983 6.137555 4.763992
x_6  4.327867 3.788407 4.102882 3.842954 4.815130 0.000000 2.793542 3.936627 5.475425 3.022680
x_7  4.287775 4.383991 3.531970 3.667172 3.465038 2.793542 0.000000 4.075392 5.251397 4.010323
x_8  6.205355 5.770031 3.916854 5.513104 5.917983 3.936627 4.075392 0.000000 5.510953 5.151613
x_9  6.197274 7.113060 6.470266 5.176213 6.137555 5.475425 5.251397 5.510953 0.000000 6.167744
x_10 2.180997 2.830447 3.733713 3.294032 4.763992 3.022680 4.010323 5.151613 6.167744 0.000000

I would like to use the nearest_neighbor_search function of this package to get the closest 3 rows for each row (ideally except the row itself). I read the documentation but am not 100% sure how to use it. Thanks.

So for the first row X_1 this code:

x <- as.data.frame(distance_columns(distances, 1))
x <- tibble::rownames_to_column(x, "id") 
x <- x[order(-x[, 2], decreasing = TRUE),]
x

produces:

     id      x_1
1   x_1 0.000000
10 x_10 2.180997
4   x_4 3.643060
2   x_2 3.843183
3   x_3 4.093911
7   x_7 4.287775
6   x_6 4.327867
5   x_5 4.935399
9   x_9 6.197274
8   x_8 6.205355

The top 3 closest rows would be x_10, x_4 and x_2.

Upvotes: 0

Views: 47

Answers (1)

Carles
Carles

Reputation: 2829

By just writing nearest_neighbor_search(distances, k= 4) where k is the number of nearest neighbours counting itself (e.g. if you want 3 then you should write 4):

nearest_neighbor_search(distances, 4, query_indices = NULL,
                        search_indices = NULL, radius = NULL)     
x_1 x_2 x_3 x_4 x_5 x_6 x_7 x_8 x_9 x_10
[1,]   1   2   3   4   5   6   7   8   9   10
[2,]  10  10   7  10   7   7   6   3   4    1
[3,]   4   6   4   3   4  10   5   6   7    2
[4,]   2   1  10   1   3   2   3   7   6    6

you get the 3 nearest neighbors for x_1. That is 1 (itself) 10,4 and 2. You can remove the first one.

Upvotes: 1

Related Questions