I Know Your Intent

With the booming of smart home market, intelligent Internet Things (IoT) devices have been increasingly involved in life. To improve user experience homes, some prior works explored how to use machine learning for predicting interactions between users and devices. However, existing solutions inferior User Device Interaction (UDI) prediction accuracy, as they ignore three key factors: routine, intent multi-level periodicity human behaviors. In this paper, we present SmartUDI, a novel accurate UDI approach homes. First, propose Message-Passing-based Routine Extraction (MPRE) algorithm mine routine behaviors, then contrastive loss is applied narrow representations among behaviors from same routines alienate different routines. Second, an Intent-aware Capsule Graph Attention Network (ICGAT) encode multiple intents while considering complex transitions Third, design Cluster-based Historical Mechanism (CHAM) capture by aggregating current sequence semantically nearest historical through attention mechanism. SmartUDI can be seamlessly deployed on cloud infrastructures IoT device vendors edge nodes, enabling delivery personalized service recommendations users. Comprehensive experiments four real-world datasets show that consistently outperforms state-of-the-art baselines with more highly interpretable results.