plot_dl_map: Plot deep learning SDM map

In boyanangelov/sdmbench: Benchmark Species Distribution Models

Description

A function that enables the plotting of deep learning predictions on a map.

Usage

plot_dl_map(raster_data, keras_model, custom_fun, map_type = "static")

Arguments

raster_data	A raster dataset containing the occurrence data.
keras_model	A trained deep learning model.
custom_fun	A custom predict function.
map_type	A logical indicating if the map should be static or interactive.

Value

An interactive leaflet map, showing the species distribution.

Examples

## Not run: 
# download benchmarking data
benchmarking_data <- get_benchmarking_data("Lynx lynx",
                                           limit = 1500)

# transform benchmarking data into a format suitable for deep learning
# if you have previously used a partitioning method you should specify it here
benchmarking_data_dl <- prepare_dl_data(input_data = benchmarking_data$df_data,
                                       partitioning_type = "default")

# perform sanity check on the transformed dataset
# for the training set
head(benchmarking_data_dl$train_tbl)
table(benchmarking_data_dl$y_train_vec)

# for the test set
head(benchmarking_data_dl$test_tbl)
table(benchmarking_data_dl$y_test_vec)

# train neural network
keras_results <- train_dl(benchmarking_data_dl)

# this function is needed for plotting
temp_fun <- function(model, input_data) {
  input_data <- tibble::as_tibble(input_data)
  data <- recipes::bake(benchmarking_data_dl$rec_obj, new_data = input_data)
  v <- keras::predict_proba(object = model, x = as.matrix(data))
  as.vector(v)
}

# plot SDM map of neural network predictions
# change the map_type argument if you want a dynamic leaflet map
plot_dl_map(benchmarking_data$raster_data$climate_variables,
           keras_results$model,
           custom_fun = temp_fun,
           map_type = "static")

## End(Not run)

boyanangelov/sdmbench documentation built on Dec. 14, 2020, 1:08 a.m.