Dynamic update upper bounds influence maximization algorithm

Junying Shang, Kan Li

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

In the field of social network analysis, influence maximization is an important research area. The goal is to find K nodes that maximize the final range of influence based on a given spread model. In the study of influence maximization, the research method based on greedy algorithm mainly focuses on how to further reduce Monte-Carlo simulations. In this paper, based on the dynamic update of the upper bound of marginal benefit of node, we propose an Upper Bound Update Greedy (UBUG) algorithm and an Upper Bound Update Heuristic (UBUH) algorithm. Experimental results show that the UBUG algorithm can reduce more MC simulations than the CELF [1] and UBLF [2] algorithms without losing the accuracy of the results. And compared to other heuristic algorithms, the UBUH not only meets the requirements of high calculation speed but also has a good accuracy.

Original languageEnglish
Title of host publicationProceedings of the 2018 2nd International Conference on Computer Science and Artificial Intelligence, CSAI 2018 - 2018 the 10th International Conference on Information and Multimedia Technology, ICIMT 2018
PublisherAssociation for Computing Machinery
Pages212-217
Number of pages6
ISBN (Electronic)9781450366069
DOIs
Publication statusPublished - 8 Dec 2018
Event2nd International Conference on Computer Science and Artificial Intelligence, CSAI 2018 - Shenzhen, China
Duration: 8 Dec 201810 Dec 2018

Publication series

NameACM International Conference Proceeding Series

Conference

Conference2nd International Conference on Computer Science and Artificial Intelligence, CSAI 2018
Country/TerritoryChina
CityShenzhen
Period8/12/1810/12/18

Keywords

  • Influence maximization
  • Social network
  • Upper bounds

Fingerprint

Dive into the research topics of 'Dynamic update upper bounds influence maximization algorithm'. Together they form a unique fingerprint.

Cite this