README.md

In scatteR: Generate Instance Space Based on Scagnostics

scatteR

scatteR generates scatterplots based on scagnostic measurements. The current implementation uses Simulated Annealing based on the GenSA package for optimization and rJava is required for the scagnostics measurement calculation.

What are these scagnostics?

Simply put scagnostics are like diagnostics for scatterplots. Each scatterplot will have a certain set of characteristics that scagnostics will show to you. You can learn more about it through this paper.

library(palmerpenguins)
library(scagnostics)
library(ggplot2)
qplot(x = bill_length_mm,y = bill_depth_mm,data=penguins)+
  theme_minimal()+
  labs(x = "Bill length",y = "Bill depth",
       title = "Scatterplot of bill length and bill depth",
       subtitle = "Data provided by palmerpenguins dataset")

The above scatterplot has the following characteristics according to scagnostics.

scagnostics(penguins$bill_length_mm,penguins$bill_depth_mm)
#>   Outlying     Skewed     Clumpy     Sparse   Striated     Convex     Skinny 
#> 0.12472358 0.74153837 0.03680493 0.04861830 0.06785714 0.56068995 0.49944162 
#>    Stringy  Monotonic 
#> 0.37671468 0.06405996 
#> attr(,"class")
#> [1] "scagnostics"

Installation

You can install the released version of scatteR from Github with:

install.packages("devtools")
devtools::install_github("janithwanni/scatteR")

Example

Simple usage

library(scatteR)
## basic example code
df <- scatteR(measurements = c("Monotonic" = 0.9),n_points = 100)
#> [1] "Epoch 1"
#> It: 1, obj value: 0.002060606061
#> [1] "Epoch 2"
#> It: 1, obj value: 0.002382925918
#> It: 20, obj value: 0.001763206699
#> It: 94, obj value: 0.000693521734
#> [1] "Epoch 3"
#> It: 1, obj value: 0.0009634972381
#> [1] "Epoch 4"
#> It: 1, obj value: 0.001210817323
#> [1] "Epoch 5"
#> It: 1, obj value: 0.0006452760813
#> [1] "Epoch 6"
#> It: 1, obj value: 0.0002109779468
#> [1] "Epoch 7"
#> It: 1, obj value: 4.33048594e-05
#> [1] "Epoch 8"
#> It: 1, obj value: 2.879774646e-05
#> [1] "Epoch 9"
#> It: 1, obj value: 1.78457193e-05
#> [1] "Epoch 10"
#> It: 1, obj value: 5.342541186e-05
#> [1] "Epoch 11"
#> It: 1, obj value: 0.0007308666196
#> [1] "Epoch 12"
#> It: 1, obj value: 0.003797876988
#> It: 27, obj value: 0.0001279653614
#> [1] "Epoch 13"
#> [1] "Epoch 14"
#> [1] "Epoch 15"
#> [1] "Epoch 16"
#> It: 1, obj value: 0.0001079452511
#> [1] "Epoch 17"
#> It: 1, obj value: 0.0001488984651
#> [1] "Epoch 18"
#> It: 1, obj value: 3.746684033e-05
#> [1] "Epoch 19"
#> It: 1, obj value: 4.910409006e-05
#> [1] "Epoch 20"
#> It: 1, obj value: 5.739727661e-06
#> [1] "Epoch 21"
#> It: 1, obj value: 2.171413779e-05
#> [1] "Epoch 22"
#> [1] "Epoch 23"
#> It: 1, obj value: 0.002067656372
#> It: 10, obj value: 6.65756155e-06
#> [1] "Epoch 24"
#> [1] "Epoch 25"
#> It: 1, obj value: 2.006914323e-05

scagnostics(df)
#>  Outlying    Skewed    Clumpy    Sparse  Striated    Convex    Skinny   Stringy 
#> 0.3374091 0.8469671 0.2213065 0.1606455 0.1296296 0.2339855 0.6467435 0.4519979 
#> Monotonic 
#> 0.8999799 
#> attr(,"class")
#> [1] "scagnostics"

plot(df$x,df$y)

Integration into the tidy workflow

library(tidyverse)
scatteR(c("Convex" = 0.9),n_points = 250,verbose=FALSE) %>% # data generation
  mutate(label = ifelse(y > x,"Upper","Lower")) %>% # data preprocessing
  ggplot(aes(x = x,y = y,color=label))+
  geom_point()+
  theme_minimal()+
  theme(legend.position = "bottom")

Using scagnostics output to generate data

generated <- scatteR(scagnostics(penguins$bill_length_mm,
                                 penguins$bill_depth_mm),
        n_points = length(penguins$bill_length_mm),verbose=FALSE)
penguins %>% 
  select(bill_length_mm,bill_depth_mm) %>% 
  drop_na() %>% 
  rename(x = bill_length_mm,y = bill_depth_mm) %>% 
  mutate(x = (x - min(x)) / (max(x) - min(x)),
         y = (y - min(y)) / (max(y) - min(y)),
         source = "penguins") %>% 
  bind_rows(generated %>% mutate(source = "generated")) %>% 
ggplot(aes(x = x,y = y,color=source))+
  geom_point()+
  theme_minimal()+
  theme(legend.position = "bottom")

scatteR documentation built on Aug. 17, 2022, 9:05 a.m.