← All projects

Forecasting via Learned Economic Representations

Led by Lalit Boyapati and Prajwal Shah

Team: Andrew Boova, Sanapala Devkinandan, and William Zhang

  • Macroeconomics
  • Representation Learning
  • Probabilistic Forecasting

Goal

Economic indicators reflect both shared conditions and local dynamics. Can a model learn from related economic histories to improve forecasts across indicators, regions, and horizons?

This project has three goals:

  1. Learn shared economic patterns while retaining indicator- and region-specific behavior.
  2. Test whether these representations improve forecast accuracy and uncertainty estimates over simpler models.
  3. Investigate whether documents provide useful predictive information beyond numerical histories.

Methodology

Coming Soon

Progress

Coming Soon