HomeGeoML & Predictive Spatial Analytics
PREDICTIVE SPATIAL ANALYTICS

Move from mapped patterns to
testable predictions.

GeoLayers GeoML adds browser-based predictive modelling to ordinary project layers, with spatial validation and outputs that remain usable across the wider platform.

Launch GeoLayers View pricing
Browser-based GISGeoAI + spatial analyticsSouth African context
PREDICTIVE SPATIAL ANALYTICS

Build models that remain connected to geography

GeoML turns project attributes and geometry into transparent predictive workflows. It is designed for practical spatial regression, classification, clustering and suitability analysis where the resulting predictions need to return to the map.

R²

Regression

Estimate continuous outcomes and inspect R², RMSE and MAE on held-out data.

C

Classification

Train multiclass models, review accuracy, Macro F1 and confusion matrices, and map predicted classes or probabilities.

K

Clustering

Use K-means to identify groups in the selected feature space and publish cluster assignments as project attributes.

◎

Suitability

Combine multiple criteria with weights and higher/lower/target-value preferences to create a 0–100 suitability score.

MODEL WORKFLOW

From project fields to prediction map

GeoML is structured so model preparation, validation and publishing are visible parts of the workflow rather than hidden implementation details.

Choose the analytical targetSelect the layer, target field and candidate numeric/categorical predictors relevant to the question.
Prepare predictorsStandardise selected variables and optionally include longitude/latitude as spatial predictors where appropriate.
Train and validateUse random holdout or spatial-block validation to test how well the model performs beyond the training observations.
Publish predictionsWrite prediction, probability, cluster or suitability fields back to a GeoJSON layer that can be mapped and reviewed in GeoLayers.
MODEL TRANSPARENCY

Metrics before maps

A visually convincing prediction surface is not enough. GeoLayers surfaces model diagnostics so users can judge whether the output deserves confidence.

  • Regression diagnostics: R², RMSE and MAE.
  • Classification diagnostics: accuracy, Macro F1 and confusion matrix.
  • Spatial-block validation to reduce overly optimistic results caused by nearby train/test observations.
  • Feature influence/importance views for explainable interpretation.
  • Exportable model configuration and mapped prediction outputs.
  • Clear separation between predictive evidence and professional or policy decisions.
PRACTICAL USE CASES

Where GeoML fits into a GIS project

Predictive spatial analysis can support prioritisation, screening and exploration across many sectors.

⌖

Site and demand modelling

Estimate continuous demand, scores or values from project attributes and location context.

▦

Spatial classification

Classify assets, land parcels or observations into operational or analytical groups.

◇

Pattern discovery

Identify clusters where natural groupings matter more than a predefined target.

✓

Multi-criteria screening

Generate transparent suitability scores from weighted, direction-aware criteria.

Build a prediction workflow in GeoLayers

Start with a project layer that contains a clear target or decision criterion and test the model before you trust the map.

Launch GeoLayers →