COMMUNITY DETECTION ON SOCIAL NETWORK USING COMMUNITY DIFFUSION WITH SOCIAL INFLUENCE SIMILARITY

• Main Author: Setiajati, Ardiansyah
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49480
• Institution: Institut Teknologi Bandung

Social networks are one of many data sources that can describe the relationship between members of the network. There are so many information that can be obtained from this relationship, one of that is community. Generally, the algorithm for community detection requires prior knowledge about underlying network topology. But to get the entire underlying network topology is not always available to retrieve. There has been a proposed method for community detection without using the underlying network topology. It is based on the diffusion information process (RCoDi). The diffusion information process will generate another structure that simpler rather than general network structure previously. Furthermore, there is another proposed method for community detection by using social influence. Social influence works by looking the influence of activities towards the relationship between members of the network. However, this method still needs an underlying network topology for detecting communities. The proposed method in this article (SICoDi) builds a community detection process based on the process of information diffusion but uses criteria from social influence. SICoDi allows detection of communities using social influence without requiring underlying network topology, but rather by utilizing the process of information diffusion. By adding new criteria, SICoDi will provide different community results compared to the RCoDi method. Based on the test results using the Normalized Mutual Information (NMI) value, SICoDi provides community results that are more optimal than RCoDi, but with large dataset sizes (nodes and cascades). However, SICoDi requires a longer execution duration than the RCoDi method because it requires more complex computation in calculating the value of Social Influence Similarity.