Skip to content

Latest commit

 

History

History
36 lines (24 loc) · 1.27 KB

todo.md

File metadata and controls

36 lines (24 loc) · 1.27 KB

Next Plans:

  • Can we actually predict rank? Using a small set of stocks conduct a study to see if its even possible to train on the parms we have and get matching rank value. Even if we overfit.
    • It's looking grim. Our MSE on the eval set is a whole number lol
    • Gonna review the columns and see if they can be simplified in some way and that they are all scaled correctly
    • Can try to train on a larger set

Other Experiments:

  • Ablation Study

    • Drop columns?
    • Using the small model to determine impact if any?
    • lstm only?

  • We never validated that the N day diff from moving avg was helpful - how do we validate that?

  • Back to reinforcement learning - what can we do here?

Longer Term:

  • Long term we should think about how to measure optimal with risk

    • Ie, ranks right now don't take into account risk, just return

  • how to find neglected stocks? Low volume, lack of index and etf inclusion

    • Get our data feed to include a list of ALL tickers on US indexes
    • Devise algo for "overlooked"

  • alfred could look at full universes soon

  • Looking at "the spreadsheet from our course" can we use that input as training data and come up with optimal quadrant?

Research: