CODEWITHSUNDEEP
rperformancesearchnumbers
CodeNoob
CodeNoob

Reputation: 1840

R: Get x neighbours from a dataframe based on column value

I wrote a function which I provide with a number which then gives me x rows having a numerical value close to the input number.

For example this is the dataframe:

test.data <- data.frame(
  number = c(0,1,3,4,6,2,7,1,3,3,4,0,1,6), 
  letter = letters[1:14]
)

Then I wrote this function to give me the neighbors:

# library(dplyr)
get.closest <- function( input.number, n.closest, data ) {
  data %>% 
    mutate(abs.score.dif = abs(input.number - number)) %>%
    arrange(abs.score.dif) %>%
    head(n.closest)
}

So for example get.closest(6, 3, test.data) will give me:

  number letter abs.score.dif rel.score.dif
1      6      e             0             0
2      6      n             0             0
3      7      g             1            -1

However I have to do this for > 20.000 numbers and my data frame is around 20.000 rows as well, making this really slow. How can this be done faster?

Upvotes: 0

Views: 312

Answers (1)

d.b
d.b

Reputation: 32548

N = 6
n = 3
df_out = transform(test.data[head(order(abs(N - test.data$number)), n),],
                   abs.diff = abs(N - number),
                   rel.diff = N - number)
df_out
#   number letter abs.diff rel.diff
#5       6      e        0        0
#14      6      n        0        0
#7       7      g        1       -1

Seems to be fast with following data

#DATA
set.seed(42)
test.data = data.frame(number = sample(0:10, 200000, TRUE),
                       letter = sample(letters, 200000, TRUE))

Upvotes: 1

Related Questions