Multivariate Probabilistic CRPS Learning with an Application to Day-Ahead Electricity Prices, 2023, Jonathan Berrisch & Florian Ziel
We apply CRPS Learning to multivariate data. We extend the smoothing methods proposed in the original CRPS Learning paper and we apply the methodology to probabilistic forecasts of multivariate day-ahead electricity prices.
Modeling volatility and dependence of European carbon and energy prices, 2023, Jonathan Berrisch & Sven Pappert & Antonia Arsova & Florian Ziel
We developed a VECM-GARCH-COPULA model for EUA, coal, natural gas and oil prices. We use maximum likelihood to estimate the model parameters and forecast the prices in a one-step process.
Paper: Finance Research Letters | Pre-Print
Presentations: Brown-Bag | ARGE Masterclass
An Introduction to Rcpp Modules, 2022, Jonathan Berrisch
This is a talk I gave at the UseR! 2022 Conference. It covers the essentials on how to expose C++ classes to R using Rcpp Modules.
Presentations: UseR!
High-Resolution Peak Demand Estimation Using Generalized Additive Models and Deep Neural Networks, 2022, Jonathan Berrisch & Michal Narajewski & Florian Ziel
This paper presents a method for estimating high-resolution electricity peak demand given lower resolution data.
Paper: Energy and AI | Pre-Print
Presentations: Brown-Bag | 2022 ISF | 2023 INREC | 2023 Statistische Woche
CRPS Learning, 2021, Jonathan Berrisch & Florian Ziel, Journal of Econometrics
This paper treats how online learning algorithms can be used for the pointwise combination of probabilistic forecasts.
Paper: Journal of Econometrics | Pre-Print
Presentations: 2021 Causal Inference Group, France | 2021 ISF | 2021 EPF Group | 2022 15TH RGS Doctoral Conference 2022 ICCF
Distributional modeling and forecasting of natural gas prices, 2021, Jonathan Berrisch & Florian Ziel
This paper presents price forecasting studies for two essential European natural-gas products.
Paper: Journal of Forecasting | Pre-Print
Projects
profoc - An R Package for Probabilistic Forecast Combination
This package implements methods which are proposed and discussed in CRPS Learning.
CRAN | GitHub | Documentation
dccpp - An R Package for fast computation of the distance covariance and distance correlation
The computation cost is only O(n log(n)) for the distance correlation (see Chaudhuri, Hu, 2019, arXiv, elsevier). The functions are written entirely in C++ to speed up the computation.
CRAN | GitHub | Documentation
sstudentt - A python package implementing the skewed-student-t distribution
This package implements the skewed student-t distribution in python. Parameterized as described in Wurtz et. al (2006) [1]. An implementation in R is already existent [2].
PyPi | Github | Documentation
Development
I develop most of my projects in a docker container that is publicly available for reuse and inspiration. This container makes it easy to reproduce results or develop projects like the profoc R package.
The most convenient way to start is using VS-Code. You can find instructions in the GitHub Repo. However, the bare-bone Docker Container is also available and sufficient.
The Repository is available on GitHub.
The docker container is also available on GitHub
References
.. [1] Wurtz, Y. Chalabi, and L. Luksan. Parameter estimation of arma models with garch/aparch errors. an r and splus software implementation. Journal of Statistical Software, 2006.
.. [2] R Implementation: https://www.gamlss.com/wp-content/uploads/2018/01/DistributionsForModellingLocationScaleandShape.pdf