rasclass (version 0.1.0)

classifyMlogit: Multinomial Logit Classsifier

Description

A function to classify raster data using Multinomial Logistic Regression.

Usage

classifyMlogit(object, anovatable = FALSE, splitfraction = 1)

Arguments

object

A rasclass-class object containing the data and a built formula.

anovatable

An optional logical variable specifying if anova statistics should be calculated.

splitfraction

An optional numeric argument specifying a fraction to split data.

Value

• A rasclass-class object, containing the classified raster in the predictedGrid slot.

Details

Data is classified using Multinomial Logistic Regression and the results are stored in the predictedGrid slot and the coefficients of the regression in the coefficients slot. The regression is calculated using the multinom function from the nnet package. The formula to classify the data is taken from the call slot of the input rasclass object.

If the argument anovatable is set TRUE, the Analysis of Variance (anova) table is calculated. For large datasets, this will significantly increase the runtime of the function.

The optional argument splitfraction allows splitting the data randomly into two parts with the fraction specified. The classification model will then be trained on one fraction and the other fraction will be used for prediction. This allows testing the preformance of the classification on data that was not an input for training the model.

See Also

accuracyAssessment, buildFormula, checkRasclass, classifyMLC, classifyMlogit, rasclass-class, rasclassRaster-class, readRaster, readRasterFolder, setRasclassData, writeRaster

Examples

# Load data from external folder
object <- readRasterFolder(path = "mypath", samplename = "mysample",
	filenames = c('myvar1.asc', 'myvar2.asc'))

# For this example, get data from a random dataframe
mysample <- sample(c(NA, 1, 2, 3), 20000, rep = TRUE)
red <- sample(c(NA,1:255), 20000, rep = TRUE)
green <- sample(c(NA,1:255), 20000, rep = TRUE)
blue <- sample(c(NA,1:255), 20000, rep = TRUE)
newdata <- data.frame(mysample,red,green,blue)

# Prepare object using the dataframe and specifying raster properties
object <- new('rasclass')
object <- setRasclassData(newdata, object, ncols = 100, nrows = 200,
	xllcorner = 0, yllcorner = 0, cellsize = 10, NAvalue = -9999,
	samplename = 'mysample')

# Classify and compute accuracy
object <- classifyMlogit(object, anova = TRUE)
object <- accuracyAssessment(object)

# Summarize and plot results
summary(object)
image(object@predictedGrid)