evaluate_dl: Evaluate deep learning model performance

In boyanangelov/sdmbench: Benchmark Species Distribution Models

Description

A function to evaluate a deep learning model.

Usage

evaluate_dl(model_keras, input_data)

Arguments

model_keras	Keras deep learning model.
input_data	A dataframe containing occurrence data parsed for deep learning using the prepare_dl_data function.

Value

A tibble containing the model estimates.

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 checks on the transformed data
# 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)

# inspect training results
keras_results$history

# you can also plot them
plot(keras_results$history)

# create evaluation tibble containing training results
keras_evaluation <- evaluate_dl(keras_results$model, benchmarking_data_dl)
head(keras_evaluation)

## End(Not run)

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