Trains the model ensemble on Google Earth Engine and scores an independent set of coordinates. Training data must contain absences: real ones, or pseudo-absences from [generate_pseudo_absences()]. Presence-only input is rejected with directions.
Usage
evaluate_models(
data,
predict_coords = NULL,
scale = 10,
methods = NULL,
aoi_year = NULL,
bg_ratio = NULL,
bg_replicates = TRUE,
balance_trees = TRUE,
n_trees = 100L,
min_leaf_population = 5L,
bag_fraction = 0.5,
shrinkage = 0.005,
max_nodes = 6L,
variables_per_split = NULL,
svm_type = "EPSILON_SVR",
svm_kernel = "RBF",
svm_cost = 10,
svm_gamma = 0.05,
maxent_beta = 1,
maxent_features = "auto",
knn_k = NULL,
knn_search_method = NULL,
knn_metric = NULL,
async = FALSE,
persist_classifier = FALSE,
gee_project = NULL,
options = list()
)Arguments
- data
Data frame of training records with `longitude`, `latitude`, `year` and a `present` column (1 = presence; include 0 rows to supply real absences).
- predict_coords
Data frame of coordinates to score (required). Include a `present` column to compute evaluation metrics on it.
- scale
Embedding resolution in metres (default 10, the native resolution).
- methods
Character vector of models to ensemble. Defaults to `c("svm", "rf", "gbt")`; also accepts `maxent`, `glm` (logistic regression fitted server-side by IRLS with equal total class weights), `similarity`, `knn`, `cart`, `mindist`. MaxEnt and glm follow the regression-family recipe of Barbet-Massin et al. (2012): a large RANDOM pseudo-absence set suits them best (see `?generate_pseudo_absences`).
- aoi_year
Year of the Alpha Earth mosaic to sample (default 2023).
- bg_ratio
Optional absence:presence ratio for the balanced background pool (the methods whose registry entry declares `pool = "balanced"`). Overrides `balance_trees` when set.
- bg_replicates
Logical (default TRUE). Train the balanced-pool methods (rf, gbt, knn) on k = min(10, ceil(10000/pool size)) replicate thinned subsets of the absences and average their predictions (Barbet-Massin et al. 2012, Table 1: several runs when few pseudo-absences are used). Requires `bg_ratio` thinning to be active; methods on the full pool are never replicated.
- balance_trees
Logical (default `TRUE`). When `TRUE`, rf/gbt and knn train on a balanced 1:1 background while svm/maxent use the full background; `FALSE` gives the trees all background points.
- n_trees, min_leaf_population, bag_fraction, shrinkage, max_nodes, variables_per_split
Tree-model (rf/gbt) hyperparameters.
- svm_type, svm_kernel, svm_cost, svm_gamma
libsvm hyperparameters (default EPSILON_SVR / RBF / cost 10 / gamma 0.05).
- maxent_beta, maxent_features
MaxEnt regularisation multiplier and feature classes (`"auto"` or a combination of L/Q/H/P/T).
- knn_k
Neighbours for kNN (default 15). Also fixes the output resolution: the surface can take only `k + 1` distinct values. Raise alongside `bg_ratio`.
- knn_search_method
kNN neighbour search: `"AUTO"`, `"LINEAR_SEARCH"`, `"KD_TREE"` or `"COVER_TREE"`. Note `KD_TREE` ignores `knn_metric`.
- knn_metric
kNN distance metric: `"EUCLIDEAN"`, `"MAHALANOBIS"`, `"MANHATTAN"` or `"BRAYCURTIS"`. Only honoured for search methods that use it.
- async
Logical; use asynchronous GEE export for large prediction sets.
- persist_classifier
Logical (default `FALSE`); persist internally-persistable classifiers (RF/CART) to a temporary GEE asset before scoring.
- gee_project
Optional Earth Engine project override (normally set via [setup_gee()]).
- options
Named list of advanced options; `batch_size` sets how many coordinates are scored per Earth Engine request (default 4000).