parallelSVM-package: Parallel-voting version of Support-Vector-Machine

In parallelSVM: A Parallel-Voting Version of the Support-Vector-Machine Algorithm

Description

By sampling your data, running the Support-Vector-Machine algorithm on these samples in parallel on your own machine and letting your models vote on a prediction, we return much faster predictions than the regular Support-Vector-Machine and possibly even more accurate predictions.

Details

Package:	parallelSVM
Type:	Package
Version:	1.0
Date:	2015-02-09
License:	GPL-2

This package consists of two main functions: parallelSVM A function which allows you to create multiple Support-Vector-Machine models: one for each core you provide. It returns a list of Support-Vector-Machine models. predict: An extension of the predict function, which uses the prediction of each Support-Vector-Machine model. When probability is TRUE, it returns the average of all predictions, otherwise it returns the class most models agree upon.

Author(s)

Wannes Rosiers

Maintainer: Wannes Rosiers <[e-mail]>

See Also

This package can be regarded as a parallel extension of svm

Examples

## Not run: 
# Create your data
data(iris)
x <- subset(iris, select = -Species)
y <- iris$Species

# Create a model
model <- parallelSVM(x, y)

# Get prediction
predictions <- predict(model, x)

# Check the quality
table(predictions,y)

## End(Not run)

parallelSVM documentation built on May 2, 2019, 9:32 a.m.