Poisson Graphical Granger Causality by Minimum Message Length

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F67985807%3A_____%2F21%3A00539725" target="_blank" >RIV/67985807:_____/21:00539725 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1007/978-3-030-67658-2_30" target="_blank" >http://dx.doi.org/10.1007/978-3-030-67658-2_30</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-030-67658-2_30" target="_blank" >10.1007/978-3-030-67658-2_30</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Poisson Graphical Granger Causality by Minimum Message Length

Popis výsledku v původním jazyce

Graphical Granger models are popular models for causal inference among time series. In this paper we focus on the Poisson graphical Granger model where the time series follow Poisson distribution. We use minimum message length principle for determination of causal connections in the model. Based on the dispersion coefficient of each time series and on the initial maximum likelihood estimates of the regression coefficients, we propose a minimum message length criterion to select the subset of causally connected time series with each target time series. We propose a genetic-type algorithm to find this set. To our best knowledge, this is the first work on applying the minimum message length principle to the Poisson graphical Granger model. Common graphical Granger models are usually applied in scenarios when the number of time observations is much greater than the number of time series, normally by several orders of magnitude. In the opposite case of “short” time series, these methods often suffer from overestimation. We demonstrate in the experiments with synthetic Poisson and point process time series that our method is for short time series superior in precision to the compared causal inference methods, i.e. the heterogeneous Granger causality method, the Bayesian causal inference method using structural equation models LINGAM and the point process Granger causality.

Název v anglickém jazyce

Poisson Graphical Granger Causality by Minimum Message Length

Popis výsledku anglicky

Graphical Granger models are popular models for causal inference among time series. In this paper we focus on the Poisson graphical Granger model where the time series follow Poisson distribution. We use minimum message length principle for determination of causal connections in the model. Based on the dispersion coefficient of each time series and on the initial maximum likelihood estimates of the regression coefficients, we propose a minimum message length criterion to select the subset of causally connected time series with each target time series. We propose a genetic-type algorithm to find this set. To our best knowledge, this is the first work on applying the minimum message length principle to the Poisson graphical Granger model. Common graphical Granger models are usually applied in scenarios when the number of time observations is much greater than the number of time series, normally by several orders of magnitude. In the opposite case of “short” time series, these methods often suffer from overestimation. We demonstrate in the experiments with synthetic Poisson and point process time series that our method is for short time series superior in precision to the compared causal inference methods, i.e. the heterogeneous Granger causality method, the Bayesian causal inference method using structural equation models LINGAM and the point process Granger causality.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10102 - Applied mathematics

Projekt

<a href="/cs/project/GA19-16066S" target="_blank" >GA19-16066S: Nelineární interakce a přenos informace v komplexních systémech s extrémními událostmi</a><br>

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2021

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Machine Learning and Knowledge Discovery in Databases. Proceedings, Part 1

ISBN

978-3-030-67657-5

ISSN

0302-9743

e-ISSN

—

Počet stran výsledku

16

Strana od-do

526-541

Název nakladatele

Springer

Místo vydání

Cham

Místo konání akce

Ghent / Virtual

Datum konání akce

14. 9. 2020

Typ akce podle státní příslušnosti

EUR - Evropská akce

Kód UT WoS článku

000717522300030