pareto_opt: Find the pareto frontier

In TianshuFeng/SemiMapper: Semi-supervised Topological Analysis

Find the pareto frontier

Description

pareto_opt finds the pareto frontier of filter functions based on Shannon indices and spread measures with function psel from R package 'rPref'.

Usage

pareto_opt(res_filter, ...)

Arguments

res_filter	The data.frame of Shannon indices and spread measures under different filter functions. The first column must be names of filter functions, and the second and third columns should be the shannon indices and spread measures.
...	Additional arguments for psel.

Value

A data.frame of the names of filter functions in the pareto frontier, as well as their Shannon indices and spread measures.

Examples

filter_names <- c('Coordinate','Eccen','LInf','Ref','DTM','Gaussian','PCA')
res_filter <- data.frame(Filter = filter_names,
                        weighted_shannon = c(1.389, 1.421, 1.453, 1.158, 1.345, 1.399, 1.349),
                        spread_measure = c(2.767, 4.101, 2.607, 4.001, 4.119, 3.957, 2.034))
res <- pareto_opt(res_filter)
res

# Illustration of Pareto frontier
library(ggplot2)
res <- pareto_opt(res_filter, top = nrow(res_filter))
res$.level[res$.level>1] <- 'Others'
res$.level[res$.level==1] <- 'Frontier'
class(res) <- 'data.frame'
gp <- ggplot(res, aes(x = weighted_shannon, y = spread_measure,
                      color = factor(.level),
                      label = Filter)) +
  geom_point(size = 3) +
  geom_step(aes(group = .level), data = res[res$.level=='Frontier',],
            direction = 'vh', size = 2) +
  geom_label(aes(fill = factor(.level)), colour = 'white', fontface = 'bold') +
  labs(fill='') + labs(color='') + xlab('Shannon index') +
  ylab('Spread measure')
gp
# The red line(s) is the Pareto frontier, and all the points on the top right
# side of these line(s) are not potentially optimal.

TianshuFeng/SemiMapper documentation built on Sept. 16, 2022, 10:26 p.m.