A0508
Title: Online copula additive models for location, scale, and shape
Authors: Christian Schulz - University of Duisburg-Essen (Germany) [presenting]
Simon Hirsch - Statkraft Trading GmbH (Germany)
Florian Ziel - University of Duisburg-Essen (Germany)
Christoph Hanck - Universität Duisburg-Essen (Germany)
Abstract: Large-scale streaming data are increasingly common in modern forecasting applications, particularly in the energy and finance sectors, and have motivated the development of online learning algorithms. In many empirical settings, jointly modeling two or more response variables conditional on covariates is of substantial interest. To achieve maximal flexibility, a generalized additive copula framework is adopted that models the marginal distributions and the copula separately, rather than assuming a multivariate distribution for the responses. Existing approaches are extended by incorporating an efficient online learning algorithm with exponential forgetting, based on online coordinate descent and LASSO-type regularization. The approach is validated in a forecasting study focused on the joint prediction of oil and gas prices. The proposed algorithms are implemented in the computationally efficient \texttt{Python} package \texttt{ondil}.
