Model-based fuzzy time series clustering of conditional higher moments

TL;DR

This paper introduces a fuzzy clustering method for time series that accounts for heteroskedasticity and non-normality by using Dynamic Conditional Score models and autocorrelation-based fuzzy C-means, demonstrated on financial data.

Contribution

It presents a novel clustering approach combining DCS models with fuzzy C-means to handle complex features in time series data.

Findings

Effective clustering of financial time series with heteroskedasticity and non-normality.

Applicable to both linear and nonlinear models.

Demonstrated on simulated and real financial data.

Abstract

This paper develops a new time series clustering procedure allowing for heteroskedasticity, non-normality and model's non-linearity. At this aim, we follow a fuzzy approach. Specifically, considering a Dynamic Conditional Score (DCS) model, we propose to cluster time series according to their estimated conditional moments via the Autocorrelation-based fuzzy C-means (A-FCM) algorithm. The DCS parametric modelling is appealing because of its generality and computational feasibility. The usefulness of the proposed procedure is illustrated using an experiment with simulated data and several empirical applications with financial time series assuming both linear and nonlinear models' specification and under several assumptions about time series density function.