iucnn_predict_status: Predict IUCN Categories from Features

In azizka/IUCNN: Neural Networks to Approximate IUCN Red List Conservation Assessments

Predict IUCN Categories from Features

Description

Uses a model generated with iucnn_train_model to predict the IUCN status of Not Evaluated or Data Deficient species based on features, generated from species occurrence records with iucnn_prepare_features. These features should be of the same type as those used for training the model.

Usage

iucnn_predict_status(
  x,
  model,
  target_acc = 0,
  dropout_reps = 100,
  return_IUCN = TRUE,
  return_raw = FALSE
)

Arguments

x	a data.set, containing a column "species" with the species names, and subsequent columns with different features, in the same order as used for iucnn_train_model
model	the information on the NN model returned by iucnn_train_model
target_acc	numerical, 0-1. The target accuracy of the overall model. Species that cannot be classified with
dropout_reps	integer, (default = 100). The number of how often the predictions are to be repeated (only for dropout models). A value of 100 is recommended to capture the stochasticity of the predictions, lower values speed up the prediction time.
return_IUCN	logical. If TRUE the predicted labels are translated into the original labels. If FALSE numeric labels as used by the model are returned
return_raw	logical. If TRUE, the raw predictions of the model will be returned, which in case of MC-dropout and bnn-class models includes the class predictions across all dropout prediction reps (or MCMC reps for bnn-class). Note that setting this to TRUE will result in large output objects that can fill up the memory allocated for R and cause the program to crash.

Value

outputs an iucnn_predictions object containing the predicted labels for the input species.

Note

See vignette("Approximate_IUCN_Red_List_assessments_with_IUCNN") for a tutorial on how to run IUCNN.

Examples

## Not run: 
data("training_occ") #geographic occurrences of species with IUCN assessment
data("training_labels")# the corresponding IUCN assessments
data("prediction_occ") #occurrences from Not Evaluated species to prdict

# 1. Feature and label preparation
features <- iucnn_prepare_features(training_occ, type = "geographic") # Training features
labels_train <- iucnn_prepare_labels(training_labels, features) # Training labels
features_predict <- iucnn_prepare_features(prediction_occ,
                                          type = "geographic") # Prediction features

# 2. Model training
m1 <- iucnn_train_model(x = features, lab = labels_train, overwrite = TRUE)

# 3. Prediction
iucnn_predict_status(x = features_predict, model = m1)

## End(Not run)

azizka/IUCNN documentation built on March 29, 2024, 9:38 a.m.