Using the SAREF ontology for interoperability and machine learning in a Smart Home environment

The Horizon2020 InterConnect project, consisting of 50 European entities from eleven
countries, aims to improve interoperability in the Smart home and office domain with the
electricity sector. SAREF is an ontology created to enable interoperability between smart
devices. However, there is a lack in the literature of practical examples to implement SAREF
in a realistic setting. We validate the practical implementation of SAREF on two aspects.
Firstly we investigate the opportunities of having various data available as a SAREF
knowledge graph in a Machine Learning (ML) context. We do this by transforming tabular
data containing sensor data into SAREF and demonstrating the communication capabilities
of IoT devices when represented as such. We furthermore demonstrate opportunities for
rule-based smart home scenario handling using OWL and SWRL languages. We finally
transform elements from the SAREF knowledge graph to common ML data formats.
Secondly we explore how to increase the explainability of ML algorithms in a Smart home
context. We surveyed multiple members from the InterConnect consortium that are involved
in developing Smart home applications. We question them about the usage of ML models
and its explainability aspects. Also we explored the explainability benefits using Jupyter
notebooks for controlling and visualizing the internal steps of ML algorithms applied to Smart
home data.
We plan further experiments in an extensive pilot within the InterConnect project, consisting
of almost 200 new apartments within the Netherlands that will be outfitted with smart devices
expressing their data compatible with SAREF