Dynamic Covariance Models for Multivariate Financial Time Series

Yue Wu, Jose Miguel Hernandez-Lobato, Ghahramani Zoubin
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):558-566, 2013.

Abstract

The accurate prediction of time-changing covariances is an important problem in the modeling of multivariate financial data. However, some of the most popular models suffer from a) overfitting problems and multiple local optima, b) failure to capture shifts in market conditions and c) large computational costs. To address these problems we introduce a novel dynamic model for time-changing covariances. Over-fitting and local optima are avoided by following a Bayesian approach instead of computing point estimates. Changes in market conditions are captured by assuming a diffusion process in parameter values, and finally computationally efficient and scalable inference is performed using particle filters. Experiments with financial data show excellent performance of the proposed method with respect to current standard models.

Cite this Paper


BibTeX
@InProceedings{pmlr-v28-wu13, title = {Dynamic Covariance Models for Multivariate Financial Time Series}, author = {Wu, Yue and Miguel Hernandez-Lobato, Jose and Zoubin, Ghahramani}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {558--566}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/wu13.pdf}, url = {https://proceedings.mlr.press/v28/wu13.html}, abstract = {The accurate prediction of time-changing covariances is an important problem in the modeling of multivariate financial data. However, some of the most popular models suffer from a) overfitting problems and multiple local optima, b) failure to capture shifts in market conditions and c) large computational costs. To address these problems we introduce a novel dynamic model for time-changing covariances. Over-fitting and local optima are avoided by following a Bayesian approach instead of computing point estimates. Changes in market conditions are captured by assuming a diffusion process in parameter values, and finally computationally efficient and scalable inference is performed using particle filters. Experiments with financial data show excellent performance of the proposed method with respect to current standard models.} }
Endnote
%0 Conference Paper %T Dynamic Covariance Models for Multivariate Financial Time Series %A Yue Wu %A Jose Miguel Hernandez-Lobato %A Ghahramani Zoubin %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-wu13 %I PMLR %P 558--566 %U https://proceedings.mlr.press/v28/wu13.html %V 28 %N 3 %X The accurate prediction of time-changing covariances is an important problem in the modeling of multivariate financial data. However, some of the most popular models suffer from a) overfitting problems and multiple local optima, b) failure to capture shifts in market conditions and c) large computational costs. To address these problems we introduce a novel dynamic model for time-changing covariances. Over-fitting and local optima are avoided by following a Bayesian approach instead of computing point estimates. Changes in market conditions are captured by assuming a diffusion process in parameter values, and finally computationally efficient and scalable inference is performed using particle filters. Experiments with financial data show excellent performance of the proposed method with respect to current standard models.
RIS
TY - CPAPER TI - Dynamic Covariance Models for Multivariate Financial Time Series AU - Yue Wu AU - Jose Miguel Hernandez-Lobato AU - Ghahramani Zoubin BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-wu13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 558 EP - 566 L1 - http://proceedings.mlr.press/v28/wu13.pdf UR - https://proceedings.mlr.press/v28/wu13.html AB - The accurate prediction of time-changing covariances is an important problem in the modeling of multivariate financial data. However, some of the most popular models suffer from a) overfitting problems and multiple local optima, b) failure to capture shifts in market conditions and c) large computational costs. To address these problems we introduce a novel dynamic model for time-changing covariances. Over-fitting and local optima are avoided by following a Bayesian approach instead of computing point estimates. Changes in market conditions are captured by assuming a diffusion process in parameter values, and finally computationally efficient and scalable inference is performed using particle filters. Experiments with financial data show excellent performance of the proposed method with respect to current standard models. ER -
APA
Wu, Y., Miguel Hernandez-Lobato, J. & Zoubin, G.. (2013). Dynamic Covariance Models for Multivariate Financial Time Series. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):558-566 Available from https://proceedings.mlr.press/v28/wu13.html.

Related Material