caret - Neural Network

This guide is designed as a quick-stop reference of how to use some of the more popular machine learning R packages with vivid. In the following examples, we use the air quality data for regression and the iris data for classification.

caret - Neural Network

The caret package (short for Classification And REgression Training) in R provides a unified interface to streamline the process of creating predictive models. In the following example, we use the caret package to utilize a neural network model fit via the nnet package. As caret is catered for in vivid, there is no need for a custom predict function here.

Regression

# load data 
aq <- na.omit(airquality)

# build caret nnet model
nn <- train(Ozone ~ ., data = aq, method = "nnet", trace = FALSE, linout = TRUE, maxit = 200)

# vivid
vi <- vivi(data = aq, fit = nn, response = 'Ozone')

Heatmap

viviHeatmap(mat = vi)

Figure 1: Heatmap of a caret neural network regression fit displaying 2-way interaction strength on the off diagonal and individual variable importance on the diagonal.

PDP

pdpPairs(data = aq, 
         fit =  nn, 
         response = "Ozone", 
         nmax = 50, 
         gridSize = 4,         
         nIce = 10)

Figure 2: Generalized pairs partial dependence plot for a caret neural network regression fit.

Classification

# Load the necessary library
library("caret")


# Load the iris dataset
data(iris)
irisMod <- iris
# Convert the Species to a binary classification: Setosa or not Setosa
irisMod$Species <- as.factor(ifelse(irisMod$Species == "setosa", "setosa", "not_setosa"))


# Train a neural network model
nn <- train(Species ~ ., data = irisMod, method = "nnet", 
               trControl = trainControl(method = "cv", number = 5), 
               tuneLength = 1,
               linout = FALSE, # this is set to FALSE for classification problems
               trace = FALSE, 
               maxit = 200)


vi <- vivi(data = irisMod, fit = nn, response = 'Species')

Heatmap

viviHeatmap(mat = vi)

Figure 3: Heatmap of a caret neural network classification fit displaying 2-way interaction strength on the off diagonal and individual variable importance on the diagonal.

PDP

pdpPairs(data = irisMod, 
         fit =  nn, 
         response = "Species", 
         nmax = 500, 
         gridSize = 20,         
         nIce = 100,
         class = 'setosa')

Figure 4: Generalized pairs partial dependence plot for a caret neural network classification fit.