knn {spatialEco}

Spatial K nearest neighbor

Description

Find K nearest neighbors for two spatial objects

Usage

knn(
  y,
  x,
  k = 1,
  d = NULL,
  ids = NULL,
  weights.y = NULL,
  weights.x = NULL,
  indexes = FALSE
)

Arguments

y

Spatial sf object or coordinates matrix

x

Spatial points or polygons object or coordinates matrix

k

Number of neighbors

d

Optional search radius

ids

Optional column of ID's in x

weights.y

A vector or matrix representing covariates of y

weights.x

A vector or matrix representing covariates of x

indexes

(FALSE/TRUE) Return row indexes of x neighbors

Details

Finds nearest neighbor in x based on y and returns rownames, index and distance, If ids is NULL, rownames of x are returned. If coordinate matrix provided, columns need to be ordered [X,Y]. If a radius for d is specified than a maximum search radius is imposed. If no neighbor is found, a neighbor is not returned

You can specify weights to act as covariates for x and y. The vectors or matrices must match row dimensions with x and y as well as columns matching between weights. In other words, the covariates must match and be numeric.

Value

A data.frame with row indexes (optional), rownames, ids (optional) and distance of k

Author(s)

Jeffrey S. Evans <[e-mail]>

See Also

nn2 for details on search algorithm

Examples

if(require(sp, quietly = TRUE)) {
library(sf)
  data(meuse, package = "sp")
  meuse <- st_as_sf(meuse, coords = c("x", "y"), crs = 28992, 
                    agr = "constant")
 
# create reference and target obs
idx <- sample(1:nrow(meuse), 10) 
  pts <- meuse[idx,]
    meuse <- meuse[-idx,]
    meuse$IDS <- 1:nrow(meuse)
 
# Find 2 neighbors in meuse
( nn <- knn(pts, meuse, k=2, ids = "IDS", indexes = TRUE) )
  plot( st_geometry(pts), pch=19, main="KNN")
    plot(st_geometry(meuse[nn[,1],]), pch=19, col="red", add=TRUE)

# Using covariates (weights)
wx = as.matrix(st_drop_geometry(meuse[,1:3]))
wy = as.matrix(st_drop_geometry(pts[,1:3]))

( nn <- knn(pts, meuse, k=2, ids = "IDS", indexes = TRUE,
            weights.y=wy, weights.x=wx) )
  plot(st_geometry(pts), pch=19, main="KNN")
    plot(st_geometry(meuse[nn[,1],]), pch=19, col="red")
	  
# Using coordinate matrices
y <- st_coordinates(pts)[,1:2]
x <- st_coordinates(meuse)[,1:2]
knn(y, x, k=2)

} else { 
  cat("Please install sp package to run example", "\n")
}