CartoBoost GeoTemporalDiffusionScenarioModel

Use GeoTemporalDiffusionScenarioModel for experimental graph-wide residual scenario generation around an existing point forecast. It is not a replacement for the point forecaster and is excluded from stable model selection.

Python Example

from cartoboost.deep import GeoTemporalDiffusionScenarioModel

model = GeoTemporalDiffusionScenarioModel(
    scenario_count=64,
    diffusion_steps=2,
    shock_scale=0.6,
)
scenarios = model.generate(
    point_forecast=[[42, 35, 18], [44, 36, 19]],
    edges=[{"source": 0, "target": 1, "weight": 0.7}],
)

FlowScenarioGenerator and ConditionalResidualDiffusion are aliases.

Use When

Use this experimental model for stress scenarios around an existing point forecast on a known graph. Do not use it as the primary point forecaster.

Browser WASM Example

GeoTemporalDiffusionScenarioModel Wasm example

Runs the Rust-backed browser export on a substantial multi-route taxi panel with a held-out prediction window.

Ready to run in this page.

Validation

Report scenario mean, variance, spatial correlation, and comparison to the point forecast. Keep capability metadata visible because this is experimental.

Limitations

• Evidence is synthetic and does not establish real-world scenario calibration.
• Scenario quality depends on the supplied point forecast and graph.
• Treat outputs as sensitivity analysis, not guaranteed probability statements.