process_features: process_features

In UrbanSkv/ela-visualization:

process_features

Description

do the final feature processing. This includes 1. Removing all problems where the landscape feature calculation encountered an error 2. Removing the features affected by shifting and scaling 3. Removing correlated features 4. Normalizing the data 5. Performing PCA

Usage

process_features(
  data,
  save_prefix,
  nfun,
  errors,
  saved_wilcoxon = FALSE,
  wilcoxon_filename = "C:\\R_Code\\saved_wilcoxon.rds"
)

Arguments

data	The dataframe of landscape features to preprocess
save_prefix	prefix of the filename that will contained the saved data about which features are affected by shifting and scaling
nfun	The total number of problems in the data
errors	Indices of rows which contained errors during feature calculation
saved_wilcoxon	Should we use existing knowledge of which features are affected by shifting and scaling, or calculate it from the provided data. If FALSE, the calculated features will be saved to a file.
wilcoxon_filename	The filename containing the names of the features affected by shifting and scaling. Used when saved_wilcoxon = TRUE

Details

The function allows two modes of operation. In the first, we calculate the features affected by shifting and scaling from the provided data. In the second, we use an existing file which lists which features are affected.

In order to not overfit the data, we first calculated the affected features using only the data from the 2014 CEC problems, saved this data, and then used these affected features when plotting comparisons between CEC and GECCO problems

Value

A processed dataframe

UrbanSkv/ela-visualization documentation built on May 20, 2022, 12:09 a.m.