Skip to main content

Dynamics

Dynamics asks whether the relationship you want to model is stable or changes over time.

If you’re unsure, you can skip this question.

Options in this tool​

Mostly stable (stationary)​

The data distribution and relationships are broadly stable.
Examples: a fixed manufacturing process, stable sensor setup.

Changes over time (non-stationary)​

The data changes (drift, seasonality, policy changes, new sensors, changing user behaviour).
Examples: social media content, economic systems, climate time series, deployment feedback loops.

Why it matters​

Non-stationary settings often need:

  • time-aware validation (avoid leakage),
  • monitoring after deployment,
  • adaptation strategies or continual learning,
  • robustness and uncertainty tracking.