FACETNET FRAMEWORK DEVELOPMENT FOR COMMUNITIES AND THEIR EVOLUTIONS ANALYSIS IN HETEROGENEOUS GRAPH DATA

• Main Author: NUR OKTAVIARI - NIM: 23515010 , KARTIKA
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28398
• Institution: Institut Teknologi Bandung

<p align="justify">Many real world problems can be modeled effectively as a complex network relationships that can be described in form of graphs. FacetNet is one of the framework to perform social network analysis especially in community detection and its evolution on dynamic graph. But recentlly, FacetNet's experimental research is still limited to homogeneous graphs, whereas in the real world, communities are made up of a complex heterogeneous networks. Therefore, this thesis will discuss the development of FacetNet framework for communities and their evolution analysis in heterogeneous graph data. <br /> <br /> Based on the results of the analysis that has been done, FacetNet can be developed to conduct community analysis and its evolution on heterogeneous graph data. The related research that can help the development of FacetNet for heterogeneous graph data is GPNRankClus. There are two approaches that can be done. First, by modifying the FacetNet component by adding the heterogeneous elements from GPNRankClus or second is by modifying GPNRankClus by adding the evolutionary element of FacetNet. From both approaches, the result analysis conclution is rather different. The conclution shown that to modify FacetNet to be used in heterogeneous graph, the only modification needed is in the FacetNet initial process in forming of an adjacency matrix. The result analysis also shown that The heterogeneous adjacency matrix formed in this study has a similar structure to the adjacency matrix of a homogeneous graph. The only difference is In the heterogeneous adjacency matrix, a dictionary is needed to define a mapping process of the node ids and its type to the index on the adjacency matrix. <br /> <br /> The experiments are done by examining the results of the FacetNet framework development using synthetic homogeneous graph data found in previous FacetNet research by Y. Lin et al in 2008, the synthetic heterogeneous graph data developed in this study and the synthetic graph data found in Ma <br /> <br /> ryadi E research in 2017. The purpose of the experiment is to test the quality and functionality of the architecture that has been developed. Based on the experimental results, FacetNet developed in this research is able to maintain the quality of every existing FacetNet functionality in previous version by Y. Lin et al in 2008 and able to describe a reasonable evolutionary cluster membership result in heterogeneous graph data. <p align="justify">