R Spatial Data – Applying KNN for Spatial Data with Two Sets of Coordinates

knnrspatial statisticsspdep

I'm trying to estimate an average value of distances to K nearest neighbors. I have two sets of coordinates and I'm trying to estimate the average distance from each point in one set of coordinates to the K nearest neighbors in the other set of coordinates.

The following post is instructive in how this could be done when the neighbors are from the same set of coordinates: Simple KNN for spatial data

But in my case I have two sets representing different feature types and I would like to know the average distance between each observation of one feature type with the K nearest neighbors of the other feature type.

Any suggestions?

Best Answer

You want something like the get.knnx function from the FNN package:

  get.knnx(data, query, k=10, algorithm=c("kd_tree", "cover_tree",
               "CR", "brute"))

It returns the k points in data that are nearest to each point in query.

Related Solutions

[GIS] n R equivalent of NEAR in ESRI ArcGIS

UPDATE:

The FNN package has a function get.knnx which can compute the N-nearest neighbours in point patterns. For what you want, this should work:

nn = get.knnx(A,B,k=1)

Which should just return the nearest neighbors between the two datasets. You can also specify what nearest neighbor algorithm it will use, be it kd_tree, cover_tree, CR, or brute force.

spatstat has a crossdist function.

Description

Computes the distances between pairs of ‘things’ taken from two different datasets.

It takes two point patterns A and B as inputs, and returns the matrix whose [i,j] entry is the distance from A[i] to B[j]. To get the nearest neighbors between two datasets using crossdist:

xdistances <- crossdist(A, B)  #Get all cross distances    

nn = numeric() 
for (i in 1:nrow(A)) {   
  xdistance <- sort(xdistances[i,], partial=1)[1]   
  nn <- append(nn, xdistance)
}

I hope that helps.

[GIS] How to perform statistical test for two sets of points

Maybe you can use PAST software to add your data as two-dimensional landmark coordinates, perform general Procrustes analysis and then a Principal Component Analysis ?

Best Answer

Related Solutions

[GIS] n R equivalent of NEAR in ESRI ArcGIS

[GIS] How to perform statistical test for two sets of points

Related Question