Replication Data for Classification of Smart Home Automations

Description

Abstract (English)

Smart home platforms allow users to create automations that combine the functionality of different smart devices to support their everyday household routines. Many users seek to share these automations with the various smart home communities; however, community-shared automations are often distributed across different forums, repository-hosting sites, video-sharing sites, and personal websites. Results from search engines such as Google will give scattered results, and the websites where automations are shared typically offer poor search alternatives, rarely offering more than keyword-based search. Because of this, discovering smart home automations can often be difficult, especially for newer users. To help increase the ease of discovery and reuse, this thesis investigates how smart home automations can be systematically classified and how this classification can be applied in practice.

This study follows a Design Science Research approach and uses a dataset consisting of a total of 2,232 automations crawled from the Home Assistant Blueprints Exchange. Various methods are used to extract information in the form of keywords from the automations, which are then used as features for machine learning clustering. The resulting clusters were used as the basis to inform the design of the classification scheme, consisting of twelve user-oriented top-level categories with their own respective sub-categories, designed with a focus on functionality. To demonstrate the practical value of the schema, a querying application was developed, with an interface supporting text search, faceted filtering, category-based browsing and detailed blueprint inspection.

Clustering was evaluated through manual interpretation of perceived usefulness and using internal metrics. The interface was evaluated against its functional requirements and use cases, though no user evaluation was performed, leaving empirical usability evaluation as the most important direction for future work.