Temel Kavramlar
The paper proposes an efficient framework to identify a seed community in a social network that has the maximum influence on a user-specified target community, satisfying both structural and keyword constraints.
Özet
The paper introduces the problem of Reverse Influential Community Search (RICS) over social networks. The goal is to find a seed community that has the highest influence on a user-specified target community, while satisfying structural and keyword constraints.
The key highlights and insights are:
The RICS problem is motivated by real-world applications such as online advertising/marketing and disease spread prevention, where we need to identify influential users to target a specific group of users.
The authors propose effective pruning strategies to reduce the search space, including keyword pruning, support pruning, and influence score pruning. These strategies help filter out invalid candidate seed communities efficiently.
An offline pre-computation phase is designed to pre-calculate and index useful information, such as keyword bit vectors, distance vectors, support upper bounds, and boundary influence upper bounds. This pre-computed data is then leveraged during the online RICS query processing.
An efficient online RICS query processing algorithm is developed, which traverses the pre-computed index and applies the proposed pruning strategies to retrieve the optimal seed community.
The paper also introduces a variant problem, Relaxed Reverse Influential Community Search (R2ICS), which returns a subgraph with relaxed structural constraints but having the maximum influence on the target community.
Comprehensive experiments on real-world and synthetic social networks demonstrate the efficiency and effectiveness of the proposed RICS and R2ICS approaches under various parameter settings.