How to calculate different well-known similarity or distance measures between two vectors in R?

Question

I want to compute the similarity (distance) between two vectors:

v1 <- c(1, 0.5, 0, 0.1)
v2 <- c(0.7, 1, 0.2, 0.1)

I just want to know if a package is available for calculating different well-known similarity (distance) measures in R? For example, "Resnik", "Lin", "Rel", "Jiang",...

The implementation of these method is not hard, but I really think it must be defined in some packages in R.

After some googling I found a package "GOSemSim", which contains most measures, but it's specific to the biomedical application and I can't use them for computing the similarity between two vectors.

Ander

The code is open-sourced. – Linguistician

What is the list of similarity measures you need access to? – Arnaud

Answer 1

"proxy" is a general library for distance and similarity measures. The following methods are supported:

"Jaccard" "Kulczynski1" "Kulczynski2" "Mountford" "Fager" "Russel" "simple matching" "Hamman" "Faith"
"Tanimoto" "Dice" "Phi" "Stiles" "Michael" "Mozley" "Yule" "Yule2" "Ochiai"
"Simpson" "Braun-Blanquet" "cosine" "eJaccard" "fJaccard" "correlation" "Chi-squared" "Phi-squared" "Tschuprow"
"Cramer" "Pearson" "Gower" "Euclidean" "Mahalanobis" "Bhjattacharyya" "Manhattan" "supremum" "Minkowski"
"Canberra" "Wave" "divergence" "Kullback" "Bray" "Soergel" "Levenshtein" "Podani" "Chord"
"Geodesic" "Whittaker" "Hellinger"

Check the following example:

x <- c(1,2,3,4,5)
y <- c(4,5,6,7,8)
l <- list(x, y)
simil(l, method="cosine")

The output is a similarity matrix between the elements of "l":

      1
2     0.978232

The only problem I have is that for some methods (such as: "Jaccard"), the following error is occurred:

simil(l, method="Jaccard")
Error in n - d : 'n' is missing

Ander

Answer 2

The dist function supports via its method argument: "euclidean", "maximum", "manhattan", "canberra", "binary" or "minkowski". See ?dist

Feldt