Online Social Platforms (OSPs) such as Twitter/X, Facebook, and Instagram have redefined social interaction and information diffusion on a global scale. Understanding influence within these large and dynamic environments--whether individual or collective--remains a core challenge in social network analysis. Traditional centrality metrics from graph theory often fail to capture the complex, multilayered nature of online interactions. This thesis proposes scalable methods to measure and analyze influence in multiplex networks, integrating both sociological perspectives and computational efficiency. It focuses on several main objectives: ensuring algorithmic scalability of an existing influence metric, linking influence to community structures, and accounting for multiple types of user interactions. Two main contributions are presented: fast algorithms for efficient influence ranking in large OSPs, and a Graph Neural Network-based approach for detecting communities across multiple interaction layers. These advances enhance both theoretical understanding and practical analysis of influence and community dynamics in online social platforms.