split_random: Randomly split dataset in multiple parts

In RSSL: Implementations of Semi-Supervised Learning Approaches for Classification

Randomly split dataset in multiple parts

Description

The data.frame should start with a vector containing labels, or formula should be defined.

Usage

split_random(df, formula = NULL, splits = c(0.5, 0.5), min_class = 0)

Arguments

df	data.frame; Data frame of interest
formula	formula; Formula to indicate the outputs
splits	numeric; Probability of of assigning to each part, automatically normalized, should be >1
min_class	integer; minimum number of objects per class in each part

Value

list of data.frames

See Also

Other RSSL utilities: LearningCurveSSL(), SSLDataFrameToMatrices(), add_missinglabels_mar(), df_to_matrices(), measure_accuracy(), missing_labels(), split_dataset_ssl(), true_labels()

Examples

library(dplyr)

df <- generate2ClassGaussian(200,d=2)
dfs <- df %>% split_random(Class~.,split=c(0.5,0.3,0.2),min_class=1) 
names(dfs) <- c("Train","Validation","Test")
lapply(dfs,summary)

RSSL documentation built on May 29, 2024, 2:38 a.m.