TY - GEN
T1 - Social event magnitudes via background influences and engagement capacities and its applications
AU - Liu, Kwei Guu
AU - Liu, Jyi Shane
N1 - Publisher Copyright:
© 2019 ACM.
PY - 2019/6/26
Y1 - 2019/6/26
N2 - Outbreaks of social events can be viewed from two angles: anomalous changes of information or popular actions. Event detection algorithms focus on the former one, while the later one is measured by social event intensity, which is a rate to show how popular an action is in a social network at a given time. The rate is relatively intuitive and can give a holistic view about activity levels in a network, but its estimation isn't easy. Inspired by event detection algorithms, this study proposes an alternative measure, social event magnitude, by using the product of background influence and cooperation value. Background influence is extracted via non-backtracking matrices, and cooperation value is obtained via engagement capacities. This alternative measure does not just integrate multisource information, but also gives a holistic view about activity levels in a network. Social event magnitudes follow a long-tailed distribution; they can be visualized for changing activities and can be applied to online event detections.
AB - Outbreaks of social events can be viewed from two angles: anomalous changes of information or popular actions. Event detection algorithms focus on the former one, while the later one is measured by social event intensity, which is a rate to show how popular an action is in a social network at a given time. The rate is relatively intuitive and can give a holistic view about activity levels in a network, but its estimation isn't easy. Inspired by event detection algorithms, this study proposes an alternative measure, social event magnitude, by using the product of background influence and cooperation value. Background influence is extracted via non-backtracking matrices, and cooperation value is obtained via engagement capacities. This alternative measure does not just integrate multisource information, but also gives a holistic view about activity levels in a network. Social event magnitudes follow a long-tailed distribution; they can be visualized for changing activities and can be applied to online event detections.
KW - Engagement Capacities
KW - Non-Backtracking Matrix
KW - Online Event Detection
KW - Spectral Algorithms
UR - http://www.scopus.com/inward/record.url?scp=85123041659&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85123041659&partnerID=8YFLogxK
U2 - 10.1145/3326467.3326481
DO - 10.1145/3326467.3326481
M3 - Conference contribution
AN - SCOPUS:85123041659
T3 - ACM International Conference Proceeding Series
BT - Proceedings of the 9th International Conference on Web Intelligence, Mining and Semantics, WIMS 2019
PB - Association for Computing Machinery
T2 - 9th International Conference on Web Intelligence, Mining and Semantics, WIMS 2019
Y2 - 26 June 2019 through 28 June 2019
ER -