Skip to content

Conversation

@wes-stone
Copy link

Replace old prediction interval function create_prediction_intervals() in final_models.R with conformal prediction implementation that uses rolling windows based upon different horizons. Also adds coverage % columns in final outputs for user to add transparency of prediction intervals.

Can view some experiments here: https://github.com/wesstone12/finnts/tree/rollling-window-v3/logs
Some basic plots: https://github.com/wesstone12/finnts/tree/rollling-window-v3/plots
https://github.com/wesstone12/finnts/tree/rollling-window-v3 also has sandbox.R in main directory to get up and running quick.

Things to look into:

  • Calibration split is pretty high due to many users submitting limited data... Would be nice for this to be more dynamic? I'm sure getting good inputs is always a challenge, but not sure how to best go about this
  • Rolling windows setup, as there can be many different combinations. Ideally, we could loop over different calibration methodologies to find the best per the data distribution (shouldn't take too long). Would likely take some experimentation with many different datasets
  • In future, will need to improve prediction interval construction methodology with more better implementations + as more research comes out
  • Need to warn users about how you can't simply just aggregate calibrated outputs (e.g., pivoting monthly data to quarterly), as they will need to re-calibrate
  • If coverage is low, prompt user to look more into data inputs
  • Plots from https://github.com/wesstone12/finnts/tree/rollling-window-v3/plots, like https://github.com/wesstone12/finnts/blob/rollling-window-v3/plots/final_forecast_M2_arimax--local--R1_meanf--local--R1.png may be interesting to enable a check box if a user wants visualizations in a separate folder

@wes-stone
Copy link
Author

@wesstone12 please read the following Contributor License Agreement(CLA). If you agree with the CLA, please reply with the following information.

@microsoft-github-policy-service agree [company="{your company}"]

Options:

  • (default - no company specified) I have sole ownership of intellectual property rights to my Submissions and I am not making Submissions in the course of work for my employer.
@microsoft-github-policy-service agree
  • (when company given) I am making Submissions in the course of work for my employer (or my employer has intellectual property rights in my Submissions by contract or applicable law). I have permission from my employer to make Submissions and enter into this Agreement on behalf of my employer. By signing below, the defined term “You” includes me and my employer.
@microsoft-github-policy-service agree company="Microsoft"

Contributor License Agreement

@microsoft-github-policy-service agree company="Microsoft"

@wes-stone wes-stone changed the title Rollling window v4 Add conformal prediction implementation Sep 5, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant