Add wasserstein-btc to Time Series Analysis (#394)

Distributional crypto-return forecasting via geodesic extrapolation on
the 2-Wasserstein manifold, with walk-forward CRPS evaluation against
both classical and named-econometric baseline panels.

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Accursed Galaxy | Robin
2026-05-30 15:43:23 +02:00
committed by GitHub
parent 12f1fdd75a
commit 8831d7e765
+1
View File
@@ -351,6 +351,7 @@ A curated list of insanely awesome libraries, packages and resources for Quants
- [gluon-ts](https://github.com/awslabs/gluon-ts) - `Python` - vProbabilistic time series modeling in Python.
- [OmniOracle](https://github.com/cesabici-bit/omni-oracle) - `Python` - Automatic discovery of non-trivial statistical relationships across 500+ time series from FRED, World Bank, EIA, and NOAA using mutual information screening, lagged MI directional testing, and FDR correction.
- [functime](https://github.com/functime-org/functime) - `Python` - Time-series machine learning at scale. Built with Polars for embarrassingly parallel feature extraction and forecasts on panel data.
- [wasserstein-btc](https://github.com/AccursedGalaxy/wasserstein-btc) - `Python` - Distributional forecasting of crypto log-returns by tangent-space geodesic extrapolation on the 2-Wasserstein manifold (quantile-function coordinates). Walk-forward CRPS evaluation over 6.75 years across 4 assets × 3 horizons; benchmarked against classical baselines (Static / RW-Drift / HS-Bootstrap / GARCH-N / GARCH-t / GJR-GARCH-t) and a named-econometric panel (HAR-RV, CAViaR-SAV, Markov-switching Normal, FIGARCH, AR(1) Stochastic Volatility, bivariate VAR+GARCH). [Live dashboard](https://accursedgalaxy.github.io/wasserstein-btc/).
- [tseries](https://cran.r-project.org/web/packages/tseries/index.html) - `R` - Time Series Analysis and Computational Finance.
- [fGarch](https://cran.r-project.org/web/packages/fGarch/index.html) - `R` - Rmetrics - Autoregressive Conditional Heteroskedastic Modelling.
- [timeSeries](https://cran.r-project.org/web/packages/timeSeries/index.html) - `R` - Rmetrics - Financial Time Series Objects.