Résumé: Influence maximization(IM) has been extensively researched in social influence analytics, aiming to find a seed set to maximize the influence spread. Previous studies have mainly focused on maximizing the number of activated nodes and rarely considered diversity. However, in real marketing, having diverse customers can diversify risk, which in turn reduces the risk of a marketing campaign. Motivated by this, we study the IM problem considering the type diversity of activated nodes and propose a new algorithm, named IMDCS. We present a metric to measure type diversity in IM and combine the maximum number of activated nodes and diversity well. An innovative community network is constructed that measures the role of the community in the whole network, and in turn determines the capacity of the community to accommodate seeds. This work also presents a new idea of the overlapping ratio to ensure that there is minimal overlap in spreading between seeds at the community scale. Experiments show that IMDCS achieves competitive results compared with other approaches.

Langue: Anglais