Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 2.111

Identifying community structure in a multi‐relational network employing non‐negative tensor factorization and GA k‐means clustering

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

The ubiquity of social networking sites leads to the generation of rich social media content. Community discovery is one of the significant tools in the analysis of social media data that is often multi‐relational due to diverse forms of user interactions. Although there has been extensive research devoted to community discovery, most of it is restricted to single‐relational networks. However, focus has been shifted to multi‐relational networks in the recent years. In this study, we aim to discover communities in the multi‐relational networks through relational learning. Our main focus is the utilization of non‐negative tensor factorization and GA k‐means clustering for community discovery. In order to incorporate the relational characteristics of the data in the learning methodology, tensors are used to model the multi‐relational network. Tensor factorization reveals the latent features of the data and shows state‐of‐the‐art results for multi‐relational learning. Once the implicit information is obtained by factorization, we apply a GA k‐means clustering algorithm for community discovery. Experiments are performed on synthetic as well as real datasets. The results obtained are quite promising and clearly demonstrate the effectiveness of our proposed scheme. WIREs Data Mining Knowl Discov 2017, 7:e1196. doi: 10.1002/widm.1196

Tensor model for multi‐relational data. Here, E1 …. En denote the entities, and R1 R m denote the relations.
[ Normal View | Magnified View ]
Comparison of the performance of various methods for YouTubeD2 for different number of clusters (a) Modularity on contact dimension (b) Modularity on co‐contact dimension
[ Normal View | Magnified View ]
Comparison of the performance of various methods for YouTubeD1 for different number of clusters (a) Modularity on contact dimension (b) Modularity on co‐contact dimension.
[ Normal View | Magnified View ]
Fitness graphs showing error bars across 100 generations for 10 runs. (a) Best fitness scores and (b) average fitness scores.
[ Normal View | Magnified View ]
An example of a multi‐relational network with 4 dimensions.
[ Normal View | Magnified View ]
The schematic view of the proposed model.
[ Normal View | Magnified View ]
Chromosome Representation and Genetic Operators (a) Chromosome Structure (b) Uniform Crossover Operator (c) Modified Whole Arithmetic Crossover Operator (d) Modified Real valued Uniform mutation operator.
[ Normal View | Magnified View ]
The Rescal Factorization.
[ Normal View | Magnified View ]
Modeling of a multi‐relational network through a three‐way tensor.
[ Normal View | Magnified View ]

Related Articles

Top Ten WIDM Articles

Browse by Topic

Technologies > Machine Learning
Algorithmic Development > Structure Discovery

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts