Influence Prediction on Social Media Network through Contents and Interaction Behaviors using Attention-based Knowledge Graph

This paper presents a model for predicting the influence of information in social media networks. Given the content, the proposed model aims to approximate the influence of one user on another by learning from both user's interaction behaviors and the vast amount of content created on the network and combining with the state-of-the-art graph convolutional and attention-based methods. We compare the performance of the proposed approach with other popular methods on one dataset, manually collected from Facebook and including the real-world interactions and contents produced by users. The experimental results show that our approach could bypass other techniques with competitive results and have more scalability for applying in real-world applications, especially in influencer and content marketing campaigns.