mlr (version 2.13)

createSpatialResamplingPlots: Create (spatial) resampling plot objects.

Description

Visualize partitioning of resample objects with spatial information.

Usage

createSpatialResamplingPlots(task = NULL, resample = NULL,
  crs = NULL, datum = 4326, repetitions = 1,
  color.train = "#0072B5", color.test = "#E18727", point.size = 0.5,
  axis.text.size = 14, x.axis.breaks = waiver(),
  y.axis.breaks = waiver())

Arguments

task

Task Task object.

resample

ResampleResult or named list with (multiple) ResampleResult As returned by resample.

crs

integer Coordinate reference system (EPSG code number) for the supplied coordinates in the Task.

datum

integer Coordinate reference system which should be used in the resulting map.

repetitions

integer Number of repetitions.

color.train

character Color for train set.

color.test

character Color for test set.

point.size

integer Point size.

axis.text.size

integer Font size of axis labels.

x.axis.breaks

numeric Custom x axis breaks

y.axis.breaks

numeric Custom y axis breaks

Value

(list of 2L containing (1) multiple `gg`` objects and (2) their corresponding labels.

CRS

The crs has to be suitable for the coordinates stored in the Task. For example, if the coordinates are UTM, crs should be set to a UTM projection. Due to a limited axis space in the resulting grid (especially on the x-axis), the data will by default projected into a lat/lon projection, specifically EPSG 4326. If other projections are desired for the resulting map, please set argument datum accordingly. This argument will be passed onto ggplot2::coord_sf.

Details

If a named list is given to resample, names will appear in the title of each fold. If multiple inputs are given to resample, these must be named.

This function makes a hard cut at five columns of the resulting gridded plot. This means if the resample object consists of folds > 5, these folds will be put into the new row.

For file saving, we recommend to use cowplot::save_plot.

When viewing the resulting plot in RStudio, margins may appear to be different than they really are. Make sure to save the file to disk and inspect the image.

When modifying axis breaks, negative values need to be used if the area is located in either the western or southern hemisphere. Use positive values for the northern and eastern hemisphere.

See Also

Other plot: plotBMRBoxplots, plotBMRRanksAsBarChart, plotBMRSummary, plotCalibration, plotCritDifferences, plotLearningCurve, plotPartialDependence, plotROCCurves, plotResiduals, plotThreshVsPerf

Examples

# NOT RUN {
rdesc = makeResampleDesc("SpRepCV", folds = 5, reps = 4)
r = resample(makeLearner("classif.qda"), spatial.task, rdesc)

##-------------------------------------------------------------
## single unnamed resample input with 5 folds and 2 repetitions
##-------------------------------------------------------------

plots = createSpatialResamplingPlots(spatial.task, r, crs = 32717,
  repetitions = 2, x.axis.breaks = c(-79.065, -79.085),
  y.axis.breaks = c(-3.970, -4))
cowplot::plot_grid(plotlist = plots[["Plots"]], ncol = 5, nrow = 2,
  labels = plots[["Labels"]])

##--------------------------------------------------------------------------
## single named resample input with 5 folds and 1 repetition and 32717 datum
##--------------------------------------------------------------------------

plots = createSpatialResamplingPlots(spatial.task, list("Resamp" = r),
  crs = 32717, datum = 32717, repetitions = 1)
cowplot::plot_grid(plotlist = plots[["Plots"]], ncol = 5, nrow = 1,
  labels = plots[["Labels"]])

##-------------------------------------------------------------
## multiple named resample inputs with 5 folds and 1 repetition
##-------------------------------------------------------------

rdesc1 = makeResampleDesc("SpRepCV", folds = 5, reps = 4)
r1 = resample(makeLearner("classif.qda"), spatial.task, rdesc1)
rdesc2 = makeResampleDesc("RepCV", folds = 5, reps = 4)
r2 = resample(makeLearner("classif.qda"), spatial.task, rdesc2)

plots = createSpatialResamplingPlots(spatial.task,
  list("SpRepCV" = r1, "RepCV" = r2), crs = 32717, repetitions = 1,
  x.axis.breaks = c(-79.055, -79.085), y.axis.breaks = c(-3.975, -4))
cowplot::plot_grid(plotlist = plots[["Plots"]], ncol = 5, nrow = 2,
  labels = plots[["Labels"]])

##-------------------------------------------------------------------------------------
## Complex arrangements of multiple named resample inputs with 5 folds and 1 repetition
##-------------------------------------------------------------------------------------

p1 <- plot_grid(plist[["Plots"]][[1]], plist[["Plots"]][[2]],
  plist[["Plots"]][[3]], ncol = 3, nrow = 1, labels = plist[["Labels"]][1:3],
  label_size = 18)
p12 <- plot_grid(plist[["Plots"]][[4]], plist[["Plots"]][[5]], ncol = 2,
   nrow = 1, labels = plist[["Labels"]][4:5], label_size = 18)

p2 <- plot_grid(plist[["Plots"]][[6]], plist[["Plots"]][[7]],
  plist[["Plots"]][[8]], ncol = 3, nrow = 1, labels = plist[["Labels"]][6:8],
  label_size = 18)
p22 <- plot_grid(plist[["Plots"]][[9]], plist[["Plots"]][[10]], ncol = 2,
  nrow = 1, labels = plist[["Labels"]][9:10], label_size = 18)

cowplot::plot_grid(p1, p12, p2, p22, ncol = 1)
# }