eval_clu: Evaluate clustering result

Description Usage Arguments Value Examples

Description

Evaluate clustering result

Usage

1
2
eval_clu(true_class, w = NULL, d = NULL, k = 10, num_clu = NULL,
  surv = NULL, type_L = c("rw", "sym", "unnormalized"), verbose = TRUE)

Arguments

true_class

A named vector of true class labels

w

affinity matrix

d

distance matrix if w is NULL, calcuate w using d

k

an integer, default 10; if w is null, w = affinity_matrix(d, k); otherwise unused.

num_clu

an integer; number of clusters; if NULL, set num_clu to be the number of classes using true_class

surv

a data.frame with at least two columns: time (days_to_death or days_to_last_follow_up), and censored (logical(1))

type_L

(parameter passed to spectral_clustering: 'type') choose one of three versions of graph Laplacian: "unnormalized": unnormalized graph Laplacian matrix (L = D - W); "rw": normalization closely related to random walk (L = I - D^(-1)*W); (default choice) "sym": normalized symmetric matrix (L = I - D^(-0.5) * W * D^(-0.5)) For more information: https://www.cs.cmu.edu/~aarti/Class/10701/readings/Luxburg06_TR.pdf

verbose

logical(1); if true, print some information

Value

a named list of size 3: "w": affinity matrix used for spectral_clustering; "clu.res": a named vector of calculated "NMI" (normalized mutual information), "ARI" (Adjusted Rand Index), and "-log10(p)" of log rank test of survival distributions of patient clusters; "labels: a numeric vector as class labels

Examples

1
2
3
4
5
6
7
8
9
library(MASS)
true.class = rep(c(1,2), each=100)
feature.mat1 = mvrnorm(100, rep(0, 20), diag(runif(20,0.2,2)))
feature.mat2 = mvrnorm(100, rep(0.5, 20), diag(runif(20,0.2,2)))
feature1 = rbind(feature.mat1, feature.mat2)
d = dist(feature1)
d = as.matrix(d)
A = affinity_matrix(d, 10)
res = eval_clu(true_class=true.class, w=A)

BeautyOfWeb/ANF documentation built on May 28, 2019, 11:57 a.m.