Encoding Individual Activation and Collective Diffusion Capability for Influence Maximization

Si Chen,Bang Wang

DOI: https://doi.org/10.1109/ijcnn60899.2024.10650712

2024-01-01

Abstract:Influence Maximization (IM) on online social networks is often modeled as an optimization problem: How to select the seed nodes to influence the most other nodes during a diffusion process? Recently some neural network-based methods have been proposed for the IM problem. Although they can outperform existing heuristic algorithms, some limitations are still yet to be addressed, such as ignoring the set collective diffusion capability and requiring huge computation to obtain training data labels. To overcome such limitations, we propose the IACD neural model for the IM problem, which consists of three modules: The first is to encode each node an individual activation capability based on its egnonet features; The second is to encode the cascade seclusion likelihood by a seed set; And the third is to encode the cascade expansion capability of a node if being included into a seed set. set. The three encoding modules are then fused to rank the most influential node to be a new seed, and the iteration proceeds till reaching the required seed number. Trained on a small synthetic network, the IACD model can be applied to large scale networks with diverse topological characteristics. Experiments are conducted on both synthetic and real-world networks, and results show that our IACD outperforms classical heuristic methods and recent neural models methods in terms of better influence capabilities in most experimented cases.