medical4pred

Silke Szymczak

The goal of medical4pred is to provide publicly available medical data sets from different resources that were consistently prepared as R data.frame. They can be used for benchmarking prediction modeling algorithms for classification and regression tasks.

Installation

You can install the development version of medical4pred from GitHub with:

# install.packages("pak")
pak::pak("imbs-hl/medical4pred")

Use case

The package contains meta information about each data set.

library(medical4pred)

data(list = "metainfo", package="medical4pred")
knitr::kable(
  metainfo,
  row.names = FALSE)

name	source	source_id	source_name	topic	other_sources	remark	no_individuals	task	freq_individuals_smallest_class	no_pred_var	no_pred_var_cat	no_pred_var_ordinal	no_pred_var_num	missing	missing_freq
bodyfat	Vanderbilt Biostatistics Datasets	NA	Body Fat Data	anthropology	Kaggle	NA	252	regression	NA	13	0	0	13	FALSE	0.0000000
BreastCancerWisconsinOriginal	UC Irvine ML Repository	15	Breast Cancer Wisconsin (Original)	oncology	NA	NA	699	classification_binary	0.3447783	9	0	9	0	TRUE	0.0022890
Dermatology	UC Irvine ML Repository	33	Dermatology	dermatology	NA	NA	366	classification_multi	0.0546448	34	1	32	1	TRUE	0.0006245
GliomaGradingClinicalandMutationFeatures	UC Irvine ML Repository	759	Glioma Grading Clinical and Mutation Features	oncology	NA	NA	839	classification_binary	0.4195471	23	22	0	1	FALSE	0.0000000
heartdisease	UC Irvine ML Repository	45	Heart Disease	cardiology	NA	NA	303	classification_multi	0.0429043	13	7	0	6	TRUE	0.0014144
HeartFailureClinicalRecords	UC Irvine ML Repository	519	Heart Failure Clinical Records	cardiology	NA	NA	299	classification_binary	0.3210702	12	5	0	7	FALSE	0.0000000
LiverDisorders	UC Irvine ML Repository	60	Liver Disorders	nephrology	NA	NA	345	regression	NA	5	0	0	5	FALSE	0.0000000
lowbirthweight	OpenML	203	lowbwt	neonatology	NA	NA	189	regression	NA	8	6	0	2	TRUE	0.0035273
NationalPollonHealthyAgingNPHA	UC Irvine ML Repository	936	National Poll on Healthy Aging (NPHA)	aging/geriatrics(?)	NA	NA	714	classification_multi	0.1834734	11	11	0	0	FALSE	0.0000000
primarytumor	OpenML	171	primary-tumor	oncology	NA	NA	339	classification_multi	0.0000000	17	17	0	0	TRUE	0.0368732

Use meta information to select data set, e.g. for a binary classification task.

data_sets_binary = subset(
  metainfo,
  task == "classification_binary",
  "name",
  drop = TRUE)
print(data_sets_binary)
#> [1] "BreastCancerWisconsinOriginal"           
#> [2] "GliomaGradingClinicalandMutationFeatures"
#> [3] "HeartFailureClinicalRecords"

Load data sets into a list (using a temporary environment).

e = new.env()
data(
  list = data_sets_binary, 
  package = "medical4pred",
  envir = e)
data_l = as.list(e)
lapply(data_l, dim)
#> $HeartFailureClinicalRecords
#> [1] 299  13
#> 
#> $BreastCancerWisconsinOriginal
#> [1] 699  10
#> 
#> $GliomaGradingClinicalandMutationFeatures
#> [1] 839  24