MPLEMENTATION AND COMPARISON OF SOCIAL INFLUENCE ANALYSIS ALGORITHMS ON SOCIAL NETWORK

• Main Author: Ilyas, Girvandi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39664
• Institution: Institut Teknologi Bandung

Social Network is a social structure made of individuals or organizations that are linked by one or more specific types of interdependency. One form of social network is Q&A network which represents relationships between individuals who asks questions and individuals who answers those questions in a certain network. The influence of these individuals to each other can be examined by using a method called social influence analysis. Social influence analysis algorithms can be used to solve influence maximization, which is a problem of finding a small subset of nodes (seed nodes) in a social network that could maximize the spread of influence. In context of Q&A network, influence maximization can be viewed as finding an expert inside a social network. To date, greedy algorithm under the linear threshold (LT) and independent cascade (IC) model achieved the best result. However, greedy algorithm has a very long computation time. To lower it, several algorithms that have been developed under the LT and IC model were LDAG and NewGreedyIC (NGIC) algorithm, respectively. In this final project, these two algorithms were implemented to solve influence maximization problem on Q&A network. To compare the two algorithms, experiments were conducted on several Q&A network. It turns out, that NGIC algorithm achieved better influence spread and faster computation time. In order to develop both algorithms, it is suggested to try implementing these algorithms in parallel to shorten the computation time. Other than that, the algorithms may be combined with another network analysis techniques such as clustering and community detection.