The Federal Reserve Bank of St. Louis is having a labor crisis; everyone in their forecasting group has taken off extended time for the holiday! Their forecasting models are usually run ad-hoc by their analysts from their own PCs, and the knowledge gap is wide enough, and the automation level low enough that no one is available to generate forecasts for their public data. They have asked us to help them throw together some rudimentary time-series forecasting software using the tool of our choosing, and guide them towards a fully-automated solution so that this never happens again.
Requirements:
-
Demonstrate forecasting proof-of-concept by providing time-series forecasts for various public economic measures (starting with just one:
MONAN, MO Total Non-Farm Employment, Thousands, Not Seasonally-Adjusted). -
No need for intermediate data persistence -- they have their own, obviously, which you'll be hitting via
GETrequests. -
Forecast results themselves should be stored in a database.
-
Go above and beyond to demonstrate the power of clean, reliable automation by providing an extra feature along with the forecasts. This extra functionality can vary between tools, as long as it's valuable & unique.