Measuring the influence and amplification of users on social network with unsupervised behaviors learning and efficient interaction-based knowledge graph

This study introduces a metric to measure the influence of users and communities on Social Media Networks. The proposed method is a combination of Knowledge Graph and Deep Learning approaches. Particularly, an effective Knowledge Graph is built to represent the interaction activities of users. Besid...

Full description

Saved in:
Bibliographic Details
Published in:Journal of combinatorial optimization Vol. 44; no. 4; pp. 2919 - 2945
Main Authors: Tran, Quan M., Nguyen, Hien D., Huynh, Tai, Nguyen, Kha V., Hoang, Suong N., Pham, Vuong T.
Format: Journal Article
Language:English
Published: New York Springer US 01.11.2022
Springer Nature B.V
Subjects:
Combinatorics
Convex and Discrete Geometry
Deep learning
Knowledge representation
Mathematical Modeling and Industrial Mathematics
Mathematics
Mathematics and Statistics
Operations Research/Decision Theory
Optimization
Social networks
Theory of Computation
Variational auto-encoder
Deep learning
Graph convolutional network
Social network analysis
Node embedding
Information propagation
Unsupervised learning
Influencer marketing
Knowledge graph
ISSN:1382-6905, 1573-2886
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study introduces a metric to measure the influence of users and communities on Social Media Networks. The proposed method is a combination of Knowledge Graph and Deep Learning approaches. Particularly, an effective Knowledge Graph is built to represent the interaction activities of users. Besides, an unsupervised deep learning model based on Variational Graph Autoencoder is also constructed to further learn and explore the behavior of users. This model is inspired by conventional Graph Convolutional layers. It is not only able to learn the attribute of users themselves but also enhanced to automatically extract and learn from the relationships among users. The model is robust to unseen data and takes no labeling effort. To ensure the state of the art and fashionable for this work, the dataset is collected by a designed crawling system. The experiments show significant performance and promising results which are competitive and outperforms some well-known Graph-convolutional-based. The proposed approach is applied to build a management system for an influencer marketing campaign, called ADVO system. The ADVO system can detect emerging influencers for a determined brand to run its campaign, and help the brand to manage its campaign. The proposed method is already applied in practice.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1382-6905
1573-2886
DOI:10.1007/s10878-021-00815-0