Clarification on 80% accuracy target — spatial unit and error metric?
Quick clarification on the 80% accuracy target for 30–60 minute forecasts — is that measured at the segment level, at specific bottlenecks, or across full origin-destination pairs? And what error metric are you using? Whether it's MAPE on travel time vs. classification accuracy on a congestion label changes the modeling approach quite a bit.
1
vote
Atharv Arya
shared this idea
-
The FCPI project does not mandate a specific evaluation metric; rather, it is left to the offeror to define the metric that best fits their proposed solution. The solicitation does not prescribe a specific mathematical error calculation approach (such as MAPE, RMSE, or classification accuracy). Proposals should include recommended metrics and calculation approaches with supporting rationale.
