Large volume of event data are becoming increasingly available in a wide
variety of applications, such as social network analysis, Internet traffic
monitoring and healthcare analytics. Event data are observed irregularly in
continuous time, and the precise time interval between two events carries a
great deal of information about the dynamics of the underlying systems. How to
detect changes in these systems as quickly as possible based on such event
data?
In this paper, we present a novel online detection algorithm for high
dimensional event data over networks. Our method is based on a likelihood ratio
test for point processes, and achieve weak signal detection by aggregating
local statistics over time and networks. We also design an online algorithm for
efficiently updating the statistics using an EM-like algorithm, and derive
highly accurate theoretical characterization of the false-alarm-rate. We
demonstrate the good performance of our algorithm via numerical examples and
real-world twitter and memetracker datasets.
%0 Generic
%1 li2016detecting
%A Li, Shuang
%A Xie, Yao
%A Farajtabar, Mehrdad
%A Song, Le
%D 2016
%K online
%T Detecting weak changes in dynamic events over networks
%U http://arxiv.org/abs/1603.08981
%X Large volume of event data are becoming increasingly available in a wide
variety of applications, such as social network analysis, Internet traffic
monitoring and healthcare analytics. Event data are observed irregularly in
continuous time, and the precise time interval between two events carries a
great deal of information about the dynamics of the underlying systems. How to
detect changes in these systems as quickly as possible based on such event
data?
In this paper, we present a novel online detection algorithm for high
dimensional event data over networks. Our method is based on a likelihood ratio
test for point processes, and achieve weak signal detection by aggregating
local statistics over time and networks. We also design an online algorithm for
efficiently updating the statistics using an EM-like algorithm, and derive
highly accurate theoretical characterization of the false-alarm-rate. We
demonstrate the good performance of our algorithm via numerical examples and
real-world twitter and memetracker datasets.
@misc{li2016detecting,
abstract = {Large volume of event data are becoming increasingly available in a wide
variety of applications, such as social network analysis, Internet traffic
monitoring and healthcare analytics. Event data are observed irregularly in
continuous time, and the precise time interval between two events carries a
great deal of information about the dynamics of the underlying systems. How to
detect changes in these systems as quickly as possible based on such event
data?
In this paper, we present a novel online detection algorithm for high
dimensional event data over networks. Our method is based on a likelihood ratio
test for point processes, and achieve weak signal detection by aggregating
local statistics over time and networks. We also design an online algorithm for
efficiently updating the statistics using an EM-like algorithm, and derive
highly accurate theoretical characterization of the false-alarm-rate. We
demonstrate the good performance of our algorithm via numerical examples and
real-world twitter and memetracker datasets.},
added-at = {2016-03-31T06:27:34.000+0200},
author = {Li, Shuang and Xie, Yao and Farajtabar, Mehrdad and Song, Le},
biburl = {https://www.bibsonomy.org/bibtex/2bb4fa8fca81398e0d7563abb5d850ca5/pixor},
description = {1603.08981v1.pdf},
interhash = {dc7454999cfa7abb79b50b011d70012b},
intrahash = {bb4fa8fca81398e0d7563abb5d850ca5},
keywords = {online},
note = {cite arxiv:1603.08981v1.pdf},
timestamp = {2016-03-31T06:27:34.000+0200},
title = {Detecting weak changes in dynamic events over networks},
url = {http://arxiv.org/abs/1603.08981},
year = 2016
}