ATOMIC: Automated Imbalanced Classification

In nunompmoniz/autoresampling: Automated Resampling Strategies for Imbalanced Learning

Description

A meta-learning-based AutoML approach to solving imbalanced classification tasks with resampling strategies.

Usage

ATOMIC(form, train, nmodels = 20, metric = "F1", numCores = 1, ...)

Arguments

form	A model formula
train	Training data
nmodels	Number of models to consider. Default is 20. Only such number of top-k models (based on internal validation performance with cross-validation methodology) will be tested.
metric	Evaluation metric used for assessing the optimisation of predictive performance. Default is F1-Score
numCores	number of cores for parallel computing
...	Other parameters

Value

A predictive model containing the workflow (algorithm+resampling strategy) that are estimated to optimise the generalisation error.

Examples

## Not run: 

library(mlbench)

data(PimaIndiansDiabetes)

form <- diabetes ~ .

atomic.m <- ATOMIC(form,PimaIndiansDiabetes)


## End(Not run)

nunompmoniz/autoresampling documentation built on April 26, 2021, 4:43 a.m.