Maximizing positive influence in competitive social networks: A trust-based solution

Abstract Online social networks provide convenience for users to propagate ideas, products, opinions, and many other items that compete with different items for influence spread. How to accurately model the spread of competitive influence is still a challenging problem. Almost all reported methods ignore the effect of trust relationships in the spread of competitive influence. Maximizing competitive influence aims to detect the top-k positive or negative influential users in social networks with competing cascades. However, finding an optimal solution to this problem is NP-hard. This study focuses on exploring the above three issues by devising a trust-based solution. First, we established a new model of trust-based competitive influence diffusion (TrCID) that simulates the spread of positive and negative influence. Second, we estimated trust values via generalized network flows and used these values to calculate influence probabilities. Finally, we developed an efficient algorithm of trust-based competitive influence maximization (TrCIM) through a heuristic pruning method. Extensive comparisons have been conducted on synthetic and real-world datasets. The effectiveness and efficiency of our approach are verified by analyzing the spread of competitive influence and the time complexity of detecting seed sets. Moreover, our approach is more practical than other baselines on real-world social networks.