My upper respiratory thing has progressed from a fever coming on to a week of roaming around feverishly while my lungs fill.
Then it got better. Then it came back. Now I’m in between.
I didn’t get as much done on MAG this weekend as I’d hoped, or during the week last week, but I did get some done to get it to a functional forecasting system.
I still have changes to make.
First, I shouldn’t be pulling symbols for configuration from SIG reports, I should be pulling from TDM derivative listings. This drops the need for pulling SIG reports for configuration, as well as ensures MAG has the configuration in place when new reports come out.
Second, I don’t need a signal forecast for NRMSE calc during configuration, that’s really only for generating the report at actual forecast time. Also, I’ll be using NRMSE as my ranking metric probably. NRMSE requires multiple days of forecasting to establish, though, and is computationally expensive, so, like directional confidence with sig, this will likely just be calculated at configuration time and rely on periodic refreshes.
I also need to build out the feature implementations.
I also need to build out model optimization and tuning.
And finally I need to work out final report generation that includes the corresponding SIG forecast data for each MAG forecast, so that the winner selection component has one less service to interface with.
These things will get MAG to a MVP state.
Now I just need to get myself to an MVP state. This cold is killin’ me. I’m dyin’. I’m a zombie.