Edge-Enabled Context-Aware Group Recommendation in Sensor-Cloud Networks: Collaboration Between Large-Scale and Lightweight Models

The rapid expansion of sensor-cloud networks and event-driven applications has accelerated the need for efficient and context-aware group recommendation services. Traditional recommender systems in event-based social networks (EBSNs) mainly target individuals, but group participation introduces unique challenges, including conflicting member preferences, cold-start issues, and contextual constraints. Meanwhile, the deployment of large-scale models (e.g., LLMs and foundation models) in cloud environments provides strong representation and reasoning capabilities but suffers from high latency and resource demands. Conversely, lightweight models deployed at the edge enable fast response but lack contextual adaptability. This paper surveys edge-enabled collaboration frameworks for context-aware group recommendation in sensor-cloud networks, emphasizing the complementary roles of large-scale and lightweight models. We categorize existing methods into aggregation strategy-based approaches and context-aware approaches, and analyze how they can be enhanced through edge-cloud collaboration. We further outline the implementation pipeline of group recommender systems in sensor-cloud environments, including data collection, edge preprocessing, cloud-based model inference, and feedback-driven adaptation. Finally, open challenges are identified, including causal modeling of group influence, online-offline behavior alignment, and dynamic decision guidance. These challenges highlight research opportunities for building scalable, interpretable, and adaptive group recommendation services in edge-enabled sensor-cloud networks, where large-scale and lightweight models work collaboratively to balance accuracy, efficiency, and responsiveness.