CAMCKG: A Framework for Trigger-Action Recommendation Combining Attention Mechanism and Continuous Kernel Graph Convolution

Trigger-Action Programming (TAP) enables the automation of Internet of Things (IoT) processes by creating rules that trigger specific actions based on certain conditions. As the number of IoT services, including those of multimedia devices and applications, continues to grow, manually selecting the right service pairs for automation becomes increasingly difficult and time-consuming. To address this, we propose CAMCKG, an innovative three-layer framework designed to recommend appropriate action services based on user-defined trigger services. Unlike existing methods, this framework effectively utilizes the complex information and relationships within TAP services to enable more intelligent automation. It introduces a novel combination of attention mechanisms and continuous kernel graph convolution to achieve better information aggregation and personalized recommendations. Experimental results demonstrate that our framework outperforms baseline models, validating its effectiveness and signifying a major advancement in service recommendation for TAP systems.