Skip to contents

Compute graph-based metrics

Source: R/getGraphMetrics.R
getGraphMetrics.Rd

Computes a selection of graph evaluation metrics using class labels.

Usage

getGraphMetrics(
  x,
  labels,
  metrics = c("SI", "NP", "AMSP", "PWC", "NCE"),
  directed = NULL,
  k = 10,
  shared = FALSE,
  level = "class",
  ...
)

Arguments

x

Either an igraph object, a list of nearest neighbors (see details below), or a data.frame or matrix (with features as columns and items as rows) from which nearest neighbors will be computed.

labels

Either a factor or a character vector indicating the true class label of each element (i.e. row or vertex) of x.

metrics

The metrics to compute. See details.

directed

Logical; whether to compute the metrics in a directed fashion. If left to NULL, conventional choices will be made per metric (adhesion, cohesion, PWC AMSP undirected, others directed).

k

The number of nearest neighbors to compute and/or use. Can be omitted if x is a graph or list of nearest neighbors.

shared

Logical; whether to use a shared nearest neighbor network instead of a nearest neighbor network. Ignored if x is not an embedding or dist object.

level

The level to calculate the metrics. Options include "element", "class" and "dataset".

...

Optional arguments for emb2knn() or emb2snn().

Value

A data.frame of metrics.

Details

The allowed values for metrics depend on the value of level:

  • If level = "element", the allowed metrics are: "SI","ISI","NP","NCE" (see below for details).

  • If level = "class", the allowed metrics are:

    • "SI": Simpson’s Index.

    • "ISI": Inverse Simpson’s Index

    • "NP": Neighborhood Purity

    • "AMSP": Adjusted Mean Shortest Path

    • "PWC": Proportion of Weakly Connected

    • "NCE": Neighborhood Class Enrichment

    • "adhesion": adhesion of a graph, is the minumum number of nodes that must be removed to split a graph.

    • "cohesion": cohesion of a graph, is the minumum number of edges that must be removed to split a graph.

  • If level = "dataset", the allowed metrics are: "SI","ISI","NP","AMSP","PWC","NCE", "adhesion","cohesion".

Examples

d1 <- mockData()
getGraphMetrics(d1[,1:2], labels=d1$class, level="class")
#>   class     SI    NP      AMSP       PWC       NCE
#> A     A 0.8024 0.844 0.1094667 0.1200000 0.3502723
#> B     B 0.5520 0.560 0.1707937 0.4666667 0.4365539