Principle:Sdv dev SDV Sequential Data Sampling
| Knowledge Sources | |
|---|---|
| Domains | Time_Series, Data_Generation |
| Last Updated | 2026-02-14 00:00 GMT |
Overview
A two-phase generation process that first samples entity contexts and then autoregressively generates time-series sequences for each entity.
Description
Sequential data sampling generates synthetic time-series by first sampling context rows (one per entity) from the fitted Gaussian copula model, then generating temporal sequences for each context using the autoregressive PARModel. The number of sequences and optionally the sequence length can be specified.
Usage
Call sample on a fitted PARSynthesizer with the desired number of sequences. Optionally specify a fixed sequence length.
Theoretical Basis
- Context generation: Sample num_sequences context rows from GaussianCopulaSynthesizer
- Sequence key assignment: Generate unique sequence key values
- Autoregressive generation: For each context row, PARModel generates timesteps one at a time
- Post-processing: Reverse-transform generated data and apply constraints
Related Pages
Implemented By
Page Connections
Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment