Generating Scenario-Centric TAP Rules for Smart Homes by Mining Historical Event Logs

Trigger-Action Programming (TAP) is a popular way of creating smart home automation applications. It can orchestrate IoT devices to fulfill user intents and make users’ daily lives more convenient. However, users’ daily lives usually have many complex scenarios that must be accomplished through several actions. The existing approaches cannot handle such situations as they mainly focus on creating simple TAP rules with a single action. This paper proposes SGen, an approach to automatically generate scenario-centric TAP rules by mining historical event traces. We first define two types of scenarios according to the characters of user activities. Accordingly, SGen identifies correlated and periodic events and uses them to synthesize scenario-centric TAP rules bottom-up without requiring all events of a potential scenario to happen at the exact moment and in the same order every time. Afterward, SGen ranks and recommends rules by prioritizing the candidates based on their diversity and significance. Finally, we evaluate SGen with two real-world datasets. The experimental results confirm that the generated scenario-centric TAP rules can match user scenarios and are more efficient in fulfilling user intents than simple rules.