In tpq/exprso: Rapid Deployment of Machine Learning Algorithms

Quick start

Welcome to the exprso GitHub page! Let's get started.

library(devtools)
devtools::install_github("tpq/exprso")
library(exprso)

library(exprso)
set.seed(1)

Importing data

To import data, we use the exprso function. This function has two arguments.

data(iris)
array <- exprso(iris[1:80, 1:4], iris[1:80, 5])

Pre-processing data

Functions with a mod prefix pre-process the data.

array <- modTransform(array)
array <- modNormalize(array, c(1, 2))

Split data

Functions with a split prefix split the data into training and test sets.

arrays <- splitSample(array, percent.include = 67)
array.train <- arrays$array.train
array.test <- arrays$array.valid

Select features

Functions with a fs prefix select features.

array.train <- fsStats(array.train, top = 0, how = "t.test")

Build models

Functions with a build prefix build models.

mach <- buildSVM(array.train,
                 top = 50,
                 kernel = "linear",
                 cost = 1)
pred <- predict(mach, array.train)
pred <- predict(mach, array.test)

calcStats(pred)

Deploy pipelines

Functions with a pl prefix deploy high-throughput learning pipelines.

pl <- plGrid(array.train,
             array.test,
             how = "buildSVM",
             top = c(2, 4),
             kernel = "linear",
             cost = 10^(-3:3),
             fold = NULL)