Parsimonious Modeling of Periodic Time Series Using Fourier and Wavelet Techniques

TL;DR

This paper introduces Fourier and wavelet techniques to create simpler, more efficient models for periodic financial time series, improving forecast accuracy compared to traditional methods.

Contribution

It presents novel parsimonious modeling approaches using Fourier and wavelet methods for periodic time series analysis, reducing parameter complexity.

Findings

More parsimonious models achieved with Fourier and wavelet techniques

Enhanced forecast efficiency demonstrated through simulations and data analysis

Outperforms traditional models like PGARCH and PACD in accuracy

Abstract

This paper proposes Fourier-based and wavelet-based techniques for analyzing periodic financial time series. Conventional models such as the periodic autoregressive conditional heteroscedastic (PGARCH) and periodic autoregressive conditional duration (PACD) often involve many parameters. The methods put forward here resulted in more parsimonious models with increased forecast efficiency. The effectiveness of these approaches is demonstrated through simulation and data analysis studies.