Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals

This study proposes a zero-inflated Rayleigh seasonal autoregressive moving average model with exogenous regressors (iRSARMAX) to model and forecast non-negative time series, accommodating the presence of zero values. The proposed iRSARMAX models the conditional mean of the continuous part of the mi...

Full description

Saved in:
Bibliographic Details
Main Authors: Aline Armanini Stefanan, Bruna G. Palm, Fabio M. Bayer
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10792885/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846118326842949632
author Aline Armanini Stefanan
Bruna G. Palm
Fabio M. Bayer
author_facet Aline Armanini Stefanan
Bruna G. Palm
Fabio M. Bayer
author_sort Aline Armanini Stefanan
collection DOAJ
description This study proposes a zero-inflated Rayleigh seasonal autoregressive moving average model with exogenous regressors (iRSARMAX) to model and forecast non-negative time series, accommodating the presence of zero values. The proposed iRSARMAX models the conditional mean of the continuous part of the mixture distribution by using a dynamic structure that considers stochastic seasonality, autoregressive and moving average terms, exogenous regressors, and a link function. It also models the mixture parameters related to the inflated (zero) values with a parsimonious dynamic structure. Furthermore, the analytical score vector was deduced and considered in the conditional maximum likelihood estimation of the introduced model parameters. The analytical Fisher information matrix was obtained and used for hypothesis testing and interval inferences for the parameters of the proposed model. Randomized quantile residuals were considered, and goodness-of-fit tests were implemented to validate the model. An extensive simulation study was performed to evaluate the performance of conditional likelihood inference over the model parameters for finite sample sizes. The proposed model excelled compared to the traditional seasonal autoregressive and moving average model and the Holt-Winters filtering in forecasting influent flow. In addition, it outperformed competitors in predicting synthetic aperture radar (SAR) image data.
format Article
id doaj-art-21b7add890fc44d8ad4edb5a3b57afb6
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-21b7add890fc44d8ad4edb5a3b57afb62024-12-18T00:01:40ZengIEEEIEEE Access2169-35362024-01-011218709918711110.1109/ACCESS.2024.351564710792885Zero-Inflated Rayleigh Dynamic Model for Non-Negative SignalsAline Armanini Stefanan0https://orcid.org/0000-0002-3994-2586Bruna G. Palm1https://orcid.org/0000-0003-0423-9927Fabio M. Bayer2https://orcid.org/0000-0002-1464-0805Postgraduate Program in Industrial Engineering, Federal University of Santa Maria, Santa Maria, State of Rio Grande do Sul, BrazilDepartment of Mathematics and Natural Sciences, Blekinge Institute of Technology, Karlskrona, Blekinge, SwedenDepartment of Statistics, Federal University of Santa Maria, Santa Maria, State of Rio Grande do Sul, BrazilThis study proposes a zero-inflated Rayleigh seasonal autoregressive moving average model with exogenous regressors (iRSARMAX) to model and forecast non-negative time series, accommodating the presence of zero values. The proposed iRSARMAX models the conditional mean of the continuous part of the mixture distribution by using a dynamic structure that considers stochastic seasonality, autoregressive and moving average terms, exogenous regressors, and a link function. It also models the mixture parameters related to the inflated (zero) values with a parsimonious dynamic structure. Furthermore, the analytical score vector was deduced and considered in the conditional maximum likelihood estimation of the introduced model parameters. The analytical Fisher information matrix was obtained and used for hypothesis testing and interval inferences for the parameters of the proposed model. Randomized quantile residuals were considered, and goodness-of-fit tests were implemented to validate the model. An extensive simulation study was performed to evaluate the performance of conditional likelihood inference over the model parameters for finite sample sizes. The proposed model excelled compared to the traditional seasonal autoregressive and moving average model and the Holt-Winters filtering in forecasting influent flow. In addition, it outperformed competitors in predicting synthetic aperture radar (SAR) image data.https://ieeexplore.ieee.org/document/10792885/ARMA modelinflated Rayleigh distributioniRSARMAX modeltime series
spellingShingle Aline Armanini Stefanan
Bruna G. Palm
Fabio M. Bayer
Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
IEEE Access
ARMA model
inflated Rayleigh distribution
iRSARMAX model
time series
title Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
title_full Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
title_fullStr Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
title_full_unstemmed Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
title_short Zero-Inflated Rayleigh Dynamic Model for Non-Negative Signals
title_sort zero inflated rayleigh dynamic model for non negative signals
topic ARMA model
inflated Rayleigh distribution
iRSARMAX model
time series
url https://ieeexplore.ieee.org/document/10792885/
work_keys_str_mv AT alinearmaninistefanan zeroinflatedrayleighdynamicmodelfornonnegativesignals
AT brunagpalm zeroinflatedrayleighdynamicmodelfornonnegativesignals
AT fabiombayer zeroinflatedrayleighdynamicmodelfornonnegativesignals