API Reference

Samplers

geovalidate.PointSampler([n_samples, ...])

Sample points uniformly at random inside a Shapely geometry.

geovalidate.ConstantClassSampler([...])

Sample exactly n_per_class points from each class.

geovalidate.StratifiedClassSampler([...])

Sample n_samples points total, allocated proportionally across classes.

geovalidate.MultinomialSampler([n_samples, ...])

Sample n_samples points via two-stage multinomial allocation.

geovalidate.PoissonSampler([n_expected, ...])

Inhomogeneous Poisson point process sampler.

Cross-validators

geovalidate.HilbertKFold([n_splits, level, ...])

Spatially balanced k-fold cross-validator via Hilbert curve ordering.

geovalidate.BallKFold([radius, n_splits])

Spatially exclusive k-fold cross-validator.

geovalidate.LeaveBallOut(radius)

Buffered leave-one-out cross-validator.

geovalidate.ClusterStratifiedKFold(clusterer)

Cluster-stratified k-fold cross-validator.

geovalidate.CellStratifiedKFold([n_splits, ...])

Stratified k-fold cross-validator using a discrete global grid system.

geovalidate.LocalBootstrap([n_bootstraps, ...])

Locally-weighted bootstrap -- resampling with spatial (or temporal) proximity weighting.

geovalidate.LocalPermutation([bandwidth, k, ...])

Spatially-constrained permutation without replacement (derangement).

Range detection

geovalidate.correlogram_range(X, y[, ...])

Estimate the distance at which spatial autocorrelation drops to zero.

geovalidate.knn_range(X, y[, max_k])

Estimate the number of nearest neighbours at which spatial autocorrelation drops to zero.

Metrics

geovalidate.area_of_applicability(X_test, ...)

Estimate the Area of Applicability (AOA) for X_test relative to X_train.

geovalidate.gearygram(X[, geometry, n_bins, ...])

Compute a Geary correlogram for univariate or multivariate spatial data.