Detection of causally anomalous time-series

Apte, Manoj; Vaishampayan, Sushodhan; Palshikar, Girish Keshav

doi:10.1007/s41060-021-00248-2

Cited by 1 publication

(1 citation statement)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, Granger Causality should be used only after such assumptions regarding the generative process for the observed data (as underlying Granger Causality) are carefully checked. Wherever possible, the causal relations obtained using Granger Causality should be compared with the known functional relations or properties [26]. [27] show that the spurious Granger Causality may be estimated if one or both TS are non-stationary.…”

Section: Recent Advancesmentioning

confidence: 99%

Causality in Time-Series: A Short Review

Palshikar,

Apte,

Vaishampayan

et al. 2023

Data Science and Intelligent Computing Techniques

View full text Add to dashboard Cite

The study of causal relations has been proved to be significant in acquisition, understanding, and representation of human knowledge across physical and biological sciences, engineering, social sciences, economics. Identifying or inferring causal relations from empirical data is a crucial step in knowledge acquisition. For study of causality, time is an important factor, as the effect will always occur after its cause, and, often, within stipulated amount of time. Thus, time-series data is considered ideal for causal inference. In this chapter we review various notions of causality in time-series data. Granger Causality is the most widely used notion for extraction of causal relations from time-series data, and has been extended to non-linear, conditional and multivariate scenarios. We also review some of the other notions of causality like Dynamic Bayesian network and Mutual Information related causal extraction techniques. We briefly touch upon Sir Austin Bradford Hill’s criteria for causality, which puts forward nine viewpoints for concluding an association relation into causality. Some applications of causality and a list of software tools are also provided. The chapter will provide sufficient information for the readers looking for a quick introduction to the field.

show abstract

Section: Recent Advancesmentioning

confidence: 99%