Data sets for Imbalanced Regression Learning

15 data sets for imbalanced regression from different domains

Contents

This repository contains the 15 data sets used in the paper:

Paula Branco, Luis Torgo, and Rita P. Ribeiro "Pre-processing Approaches for Imbalanced Distributions in Regression" submitted to Neurocomputing Journal.

The data sets are available in 3 formats: Rdata, csv and arff.

You can either download the data sets in all formats, or in one desired format.

This page allows you to select the format of the data sets and download them in the selected format.

More details on the data sets

The main characteristics of the 15 regression data sets in this folder are as follows:

ID	Data Set	N	tpred	p.nom	p.num	nRare	% Rare
DS1	a6	198	11	3	8	33	16.7
DS2	Abalone	4177	8	1	7	679	16.3
DS3	a3	198	11	3	8	32	16.2
DS4	a4	198	11	3	8	31	15.7
DS5	a1	198	11	3	8	28	14.1
DS6	a7	198	11	3	8	27	13.6
DS7	boston	506	13	0	13	65	12.8
DS8	a2	198	11	3	8	22	11.1
DS9	fuelCons	1764	37	12	25	164	9.3
DS10	heat	7400	12	4	8	664	9.0
DS11	availPwr	1802	15	7	8	157	8.7
DS12	cpuSm	8192	12	0	12	713	8.7
DS13	maxTorque	1802	32	13	19	129	7.2
DS14	bank8FM	4499	8	0	8	288	6.4
DS15	Accel	1732	14	3	11	89	5.1

where,

N represents the total number of cases;

tpred represents the number of predictors;

p.nom represents the number of nominal predictors;

p.num represents the number of numeric predictors;

nRare represents the number of cases with target variable relevance above 0.8; and

% Rare=nRare/N.

Directly importing the data sets into R

To import the .Rdata file directly into R you will need the repmis and DMwR R packages.

You can install them as follows:

install.packages(c("repmis", "DMwR"))

Then, load the libraries:

library(DMwR)
library(repmis)

And use the source_data function:

repmis::source_data(url = "https://github.com/paobranco/Imbalanced-Regression-DataSets/blob/master/RDATA_data/DataSets15.Rdata?raw=true")

This will load into R an object named DSs which contains a list with 15 objects of class dataset from package DMwR.

You can acces and explore the data sets as follows.

Observe the first task:

DSs[[1]]

In particular for this data set, you may check it's name and the formula of the task.

DSs[[1]]@name
DSs[[1]]@formula

See the first lines of the first data set:

head(DSs[[1]]@data)