buildGLMJavaClass: Build a Java class of a GLM model, creating a predict() Java...

In wtcooper/glmToJava: Simple R method to Export a Java GLM Model

Description

Build a Java class of a GLM model, creating a predict() Java method. the dataset through shiny (more control than expore_all_data() provides).

Usage

buildGLMJavaClass(modCoefs, modType = "gaussian", filePath, package,
  threshold = 0.5, addMainMeth = FALSE)

Arguments

modCoefs	A data.frame with coefficient names and values. The intercept must be passed as containing case sensitive text "Intercept". E.g. (Intercept) as from glmnet is fine.
modType	A string of the family distribution (gaussian, poisson, binomial, multinomial).
filePath	Where to write the Java class too.
package	The name of the Java class package.
threshold	For binary classification, the probability threshold to use (default=0.5).

Examples

data(iris)
library(glmnet)

### Fit Gaussian model ###
x = as.matrix(iris[,2:4])
y = iris[,1]
normFit=cv.glmnet(x,y, family="gaussian")
normCoefs = as.data.frame(t(as.matrix(coef(normFit, s="lambda.min"))))

buildGLMJavaClass(modCoefs=normCoefs, modType="gaussian",
		filePath="./NormalMod.java", package="glm2java", addMainMeth=TRUE)


### Fit Poisson model ###
poisFit=cv.glmnet(x,y, family="poisson")
poisCoefs = as.data.frame(t(as.matrix(coef(poisFit, s="lambda.min"))))

## With the default naive threshold (0.5)
buildGLMJavaClass(modCoefs=poisCoefs, modType="poisson",
		filePath="./PoissonMod.java", package="glm2java", addMainMeth=TRUE)


### Fit binomial model ###
x = as.matrix(iris[,1:4])
y = iris[,5]
y = as.character(y)
y[y=="virginica"]="versicolor"
y=as.factor(y)

# With the default naive threshold (0.5)
buildGLMJavaClass(modCoefs=binCoefs, modType="binomial",
		filePath="./BinaryMod.java", package="glm2java", addMainMeth=TRUE)


# With a user provided threshold (0.2)
buildGLMJavaClass(modCoefs=binCoefs, modType="binomial",
		filePath="./BinaryMod_2.java", package="glm2java", threshold=0.2, addMainMeth=TRUE)



### Fit multinomial model ###
x = as.matrix(iris[,1:4])
y = iris[,5]
multiFit = cv.glmnet(x,y, family="multinomial" )
multiCoefs = lapply(coef(multiFit, s="lambda.min"),
  FUN=function (x) as.data.frame(t(as.matrix(x))))

buildGLMJavaClass(modCoefs=multiCoefs, modType="multinomial",
filePath="./MultinomialMod.java", package="glm2java", addMainMeth=TRUE)

wtcooper/glmToJava documentation built on May 4, 2019, 11:59 a.m.