Dynamic Covariance Models for Multivariate Financial Time Series

TL;DR

This paper introduces a Bayesian dynamic covariance model for multivariate financial data that effectively captures market shifts, avoids overfitting, and is computationally scalable, outperforming standard models.

Contribution

A novel Bayesian dynamic covariance model using diffusion processes and particle filters for scalable, accurate financial time series analysis.

Findings

Outperforms standard models in financial data prediction

Effectively captures market condition shifts

Reduces overfitting and computational costs

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.