Changepoint detection in Gaussian sequential data with copula-based Markov chain models in an online environment

Abstract

Listen

Translate article

Online changepoint detection involves determining if there are changes in a sequential data structure over time. Addressing the dependence structure is crucial in analysing time series data. To address this, we propose a copula-based Markov chain model that uses the Clayton copula and the Gaussian distribution as the marginal distribution. To overcome the limitations of the independence-based conjugate structure, we incorporate the Markov chain Monte Carlo (MCMC) method for online changepoint detection. Simulation experiments demonstrate that our proposed model outperforms competitors in terms of the Precision, F1 Score, and Mean Absolute Error (MAE) across various scenarios. In empirical studies, we apply our model to analyse the daily log returns of the S&P 500 Index during the significant events like the 2008 financial crisis and the 2020 COVID-19 pandemic. The results successfully capture the structural changes in these real-world applications.