A Genetic Algorithm-Based Heuristic for Rumour Minimization in Social Networks

Rumours in online social networks can significantly damage the ecosystem of the society. It is important to timely identify and control the rumor spread. Rumour identification itself is a challenging problem. The rumour control comes in to play once the rumor is identified. There are mainly two kinds of rumour control strategies: 1) Network disruption strategies and 2) Truth propagation strategies. The diffusion model addresses how the rumour \(/\) truth is spreading in the network. Independent Cascade Model (ICM) and Linear Threshold Model (LTM) are the two well known diffusion models. These diffusion models address propagation of a single \(/\) independent message. Yang et al. proposed Linear Threshold model with One Direction state Transition (LT1DT), which handles simultaneous propagation of two messages (rumor and truth) which are opposite in nature. The nodes which start the spread of rumour message initially are called rumour seed nodes. Under the LT1DT model, for a given rumour seed set, the rumour minimization problem asks to find k truth seed nodes that minimize the overall rumour spread in the network. In this paper, we propose genetic algorithm framework based heuristic and a pruning technique to compute the truth seed nodes. We have implemented all the existing algorithms and the proposed heuristics. We have done an extensive experimentation on synthetic and real datasets and compared our results with existing heuristics. Proposed heuristics have shown significant improvement in minimizing the rumour spread.