R/similarity_taxon.r

In david-beauchesne/iEat: Functions to apply a machine learning algorithm to predict species trophic interactions

# Similarity matrix as a single functions
similarity_taxon <- function(S0, wt, taxa) {
  # Note: the similarity on the diagonal has to be set to 1 since it's all the same species.
  # taxa is either resource or consumer

  similarity.matrix <- matrix(nrow = nrow(S0), ncol = nrow(S0), dimnames = list(S0[, "taxon"], S0[, "taxon"]))

  taxonomy <- vector("list", nrow(S0))
  taxon <- vector("list", nrow(S0))
  for (i in 1:nrow(S0)) {
    taxonomy[[i]] <- unlist(strsplit(S0[i, "taxonomy"], " \\|\\ "))
    if (length(which(taxonomy[[i]] == "NA")) > 0) {
      taxonomy[[i]] <- taxonomy[[i]][-which(taxonomy[[i]] == "NA")]
    }

    if (taxa == "consumer") {
      taxon[[i]] <- unlist(strsplit(S0[i, "resource"], " \\|\\ "))
    } else if (taxa == "resource") {
      taxon[[i]] <- unlist(strsplit(S0[i, "consumer"], " \\|\\ "))
    }
  }

  pb <- txtProgressBar(min = 0, max = nrow(S0), style = 3)

  for (i in 1:nrow(S0)) {
    for (j in i:nrow(S0)) { # No need to evaluate both side of the matrix diagonal for the similarity matrix
      similarity.matrix[i, j] <- similarity.matrix[j, i] <- tanimoto_traits(
        resource_x = taxon[[i]],
        resource_y = taxon[[j]],
        trait_x = taxonomy[[i]],
        trait_y = taxonomy[[j]],
        trait_weight = wt
      )
    } # j
    setTxtProgressBar(pb, i)
  } # i
  close(pb)

  diag(similarity.matrix) <- 1

  return(similarity.matrix)
} # similarity_taxon function