将一列值四舍五入到另一列中最接近的值（R 代码）

Question

I have a data frame of coordinates like the one bellow.

longitude latitude
1    95.93604 41.25908
2    95.93371 41.25941
3    95.93137 41.25974
4    95.92904 41.26008
5    95.92670 41.26041

and I want to round both the longitudes and latitudes to the nearest values in a existing data frame with coordinates like the one below.

latitude longitude
41.45131  96.42024
40.81344  96.66093
41.11293 102.85215
40.37834  96.61095
42.84468  97.40045
41.18000  96.11592
40.69164  99.53231
40.37834  96.61095
41.34500  95.95712

How can I do this in R? I tried using the interp1 function from the package pracma, but I was not able to get the correct result. I suppose I could write a function of my own, but I am also curious if there is a simpler and more elegant way of going about this. If you have any other suggestions for how I can find the closest corresponding coordinates from one data frame to the other, that would be appreciated too! Thank you!

Answer 1

We can use a k-nearest neighbour classification algorithm.

The class package has the knn1 function that can do this. It returns the indices of the nearest neighbours based on two data frames, train (the data containing your "rounded" coordinates) and test (your actual data).

(ind <- knn1(train, test, cl=1:nrow(train)))
[1] 1 9 9 3 9
Levels: 1 2 3 4 5 6 7 8 9

This shows that the first row of test is nearest to the first row of train , the fourth row nearest the third row, and all other rows nearest to the last (9th) row.

We can then use these indices to extract the rounded coordinates into two new columns (or replace the existing ones).

test$longitude.rnd <- train$longitude[ind]
test$latitude.rnd <- train$latitude[ind]
test
  longitude latitude longitude.rnd latitude.rnd
1  96.42604 41.45908      96.42024     41.45131
2  95.93371 41.25941      95.95712     41.34500
3  95.93137 41.25974      95.95712     41.34500
4 102.82904 41.16008     102.85215     41.11293
5  95.92670 41.26041      95.95712     41.34500

---

Test Data : (I modified two rows to show variation, otherwise all rows would return the 9th)

test <- structure(list(longitude = c(96.42604, 95.93371, 95.93137, 102.82904, 
95.9267), latitude = c(41.45908, 41.25941, 41.25974, 41.16008, 
41.26041)), row.names = c("1", "2", "3", "4", "5"), class = "data.frame")

  longitude latitude
1  96.42604 41.45908
2  95.93371 41.25941
3  95.93137 41.25974
4 102.82904 41.16008
5  95.92670 41.26041

Train Data : (no modification, except that I swapped the columns to match the test data)

train <- structure(list(longitude = c(96.42024, 96.66093, 102.85215, 96.61095, 
97.40045, 96.11592, 99.53231, 96.61095, 95.95712), latitude = c(41.45131, 
40.81344, 41.11293, 40.37834, 42.84468, 41.18, 40.69164, 40.37834, 
41.345)), class = "data.frame", row.names = c(NA, -9L))

  longitude latitude
1  96.42024 41.45131
2  96.66093 40.81344
3 102.85215 41.11293
4  96.61095 40.37834
5  97.40045 42.84468
6  96.11592 41.18000
7  99.53231 40.69164
8  96.61095 40.37834
9  95.95712 41.34500

Answer 2

A simpler approach (A function Indeed).

#' title, Find nearest value
#'
#' @param x, element for which nearest value has to be found.
#' @param ref_col, field name from reference table from which x has to be compared.
#'
#' @return, nearest value to x.
find_nearest_fun <- function(x, ref_col = "latitude") {
  #browser()
  ref_field_vec <- ref_d[[ref_col]] # reference field 
  min_idx <- which.min(abs(ref_field_vec - x)) 
  return(ref_field_vec[min_idx])
}

d %>% mutate(nearest_lat = map_dbl(latitude, ~find_nearest_fun(.x)),
             nearest_long = map_dbl(longitude, ~find_nearest_fun(.x,ref_col = "longitude"))) %>% 
  view()

Data ( a bit modified for functionality check )

tribble(~latitude, ~longitude,
41.45131,  96.42024,
40.81344,  96.66093,
41.11293,  102.85215,
40.37834,  96.61095,
42.84468,  97.40045,
41.18000,  96.11592,
40.69164,  99.53231,
40.37834,  96.61095,
41.34500,  95.95712
) -> ref_d


tribble(~longitude ,~latitude,
95.93604  ,41.25908,
95.93371  ,41.25941,
95.93137  ,41.25974,
95.92904  ,41.26008,
95.92670  ,41.26041,
98.92670  ,40.26041,
96.92670  ,40.60412
) -> d