match_evaluate_multiple: match_evaluate_multiple

In pipeComp: pipeComp pipeline benchmarking framework

Description

Function to match cluster labels with 'true' clusters using the Hungarian algorithm, and return precision, recall, and F1 score. Written by Lukas Weber in August 2016 as part of his cytometry clustering comparison, with just slight modifications on initial handling of input arguments.

Usage

match_evaluate_multiple(clus_algorithm, clus_truth = NULL)

Arguments

clus_algorithm	cluster labels from algorithm
clus_truth	true cluster labels. If NULL, will attempt to read them from the names of 'clus_algorithm' (expecting the format 'clusterName.cellName')

Value

A list.

Examples

# random data
dat <- data.frame( 
 cluster=rep(LETTERS[1:3], each=10),
 x=c(rnorm(20, 0), rnorm(10, 1)),
 y=c(rnorm(10, 1), rnorm(20, 0))
)
# clustering
dat$predicted <- kmeans(dist(dat[,-1]),3)$cluster
# evaluation
match_evaluate_multiple(dat$predicted, dat$cluster)

pipeComp documentation built on Nov. 8, 2020, 7:35 p.m.