Evaluate a single BirdFlow model (default evaluation function)

Loads a fitted BirdFlow HDF5, computes interval-based metrics, optional route statistics (both conditional on observed tracks and unconditional synthetic routes by season), and PIT calibration diagnostics. Also extracts key hyperparameters and distribution-level correlations.

Usage

evaluate_model(bf_path, data, params)

Arguments

bf_path

Path to a fitted BirdFlow HDF5 file.

data

A list where [[1]] holds interval data and [[2]] holds one-week interval data used for PIT calibration.

params

Parameter list (usually evaluator$batch_trainer$params) providing output_path, season, and other configuration used by helpers.

Value

A list with elements:

df

A one-row tibble of summary metrics and hyperparameters.

obs

The interval data object passed in data[[1]].

metric_for_each_transition

Per-transition metrics table.

Details

Side effects:

• Saves a PIT calibration RDS to <output_path>/pit_data/.

• Prints a compact tibble of derived metrics.

See also

evaluate.BatchBirdFlowEvaluator(), BirdFlowR::calc_interval_metrics(), BirdFlowR::distribution_performance(), BirdFlowR::route()

Examples

if (FALSE) { # \dontrun{
one <- split_data$training_data
two <- split_data$training_data_one_week
out <- evaluate_model(".../amewoo_150km_xyz.hdf5", data = list(one, two), params = trainer$params)
out$df
} # }