demo/classify.R

In jr-packages/jrPredictive: Jumping Rivers: Machine Learning with R

library(shiny)
library(caret)
data("mixture.example", package = "ElemStatLearn")
x <- mixture.example$x
g <- mixture.example$y
xnew <- mixture.example$xnew
df = data.frame(x = x[, 1], z = x[, 2], y = factor(g, labels = c("No", "Yes")))
dd = data.frame(x = xnew[, 1], z = xnew[, 2])
server = function(input, output) {
  slider = reactive({
    if (input$method == "knn")
      sliderInput("k", "# Nearest Neighoburs",
                  min = 1, max = 30, step = 2, value = 10)
  })

  output$slider = renderUI({
    slider()
  })

  grid = reactive({
    if (input$method == "knn") data.frame(k = input$k)
    else NULL
  })
  modelfit = reactive({
    train(y ~ x + z, data = df, method = input$method, tuneGrid = grid(),
          trControl = trainControl(method = "none"),
          preProcess = c("center", "scale"))
  })
  output$plot = renderPlot({
    jrPredictive::boundary_plot(
      modelfit(), df$x, df$z,
      df$y, lwd = 5, jitter = FALSE
    )
  })


}

ui = fluidPage(
  titlePanel("Classification boundaries"),
  sidebarLayout(
    sidebarPanel(
      radioButtons("method", "Classifier",
                   choices = c("lda", "qda", "knn", "glm")),
      uiOutput("slider")
    ),
    mainPanel(plotOutput("plot"))
  )
)


shinyApp(server = server, ui = ui)