subtopic

Generate synthetic data (3 methods)

(1) Rule-based simulation, (2) statistical model (Gaussian copula with SDV), (3) local oversampling (SMOTE) — on the training split only. Concepts: Simulation; copula-based synthesis; CTGAN (glimpse); SMOTE as synthesis Project A: GaussianCopula on German Credit train → 1,000 synthetic applicants. Project B: Simulate a heat-wave week: shift temperatures +6 °C and generate the matching inputs for scenario forecasts.