ML Training Window Assessor
Evaluate whether and how far a model's training window can be extended by incorporating a new data source or removing a data dependency bottleneck.
Problem
Multi-output temporal models (e.g., per-term enrollment prediction) often have training windows that are shorter than they appear. A nominal date range of "10 months" may actually be 3 months for one output and 9 months for another, due to dynamic lookforward windows and label validity constraints. This lea
[Description truncada. Veja o README completo no GitHub.]