nicl-nno
commented
2 years ago The idea of basic implementation is the following:
1) Obtain set of TOP-n best pipelines from OptHistory (see 'leaderboard') 2) Estimate the quantile-based boundaries of interval for each point of forecast 3) Add interval to the visualisation of forecast.
Expected results:
- new feature for the interval forecasts
- modified visulisation
- unit test for new features
- new example (of modified old example) for interval forecasting
valer1435
When we forecast futures values, there is always uncertainty around the values that needed to be quantified. This is called risk. It would be nice if we can provide a probabilistic forecast around the forecasted values.