# Smart home environment data across 4 European countries

## TL;DR

This paper presents a dataset of smart home environmental data collected over nearly a year from 62 users in four European countries, focusing on elderly individuals.

## Contribution

The dataset introduces a long-term, real-world smart home data collection with diverse sensors and timestamps across multiple countries.

## Key findings

- Environmental data can estimate room presence without relying on direct motion sensors.
- The dataset's strength lies in its variety of sensors and long-term high-frequency measurements.
- The lack of validated ground truth limits supervised learning applications.

## Abstract

This paper describes a dataset of anonymised smart home environment data that was collected during a project over 359 days (16.05.2023-08.05.2024). The dataset contains information about temperature (°C), humidity (%), ambient light (lux), CO2 (ppm), VOC (ppm), sound pressure level (dB) in a time interval of 2–5 min in addition to event based data from PIR-Sensors and door contact sensors. Additionally, time and location information for each data point is available in the form of a time stamp, the user ID, the room and the country. The dataset was collected in 4 different European countries from a total of 62 users at their residential settings. Different installations had different sets of sensors, meaning not all parameters were measured at every location. The target group for the field trials was elderly people 65+.

During the project it could be shown that the data can be used to estimate presence in a room, based on the environmental data only, where the output of the PIR-Sensors were used as proxy labels. The weakness of the dataset is the lack of validated ground truth, which makes supervised learning approaches difficult. The strength of the dataset lies in the variety of sensors including sound pressure and the long period (nearly 1 year) of high frequency measurements in different countries.

Collecting data in real-world residential settings is challenging, but by making this dataset publicly available, we provide researchers with a valuable resource to explore smart home applications, presence detection, and environmental monitoring in everyday life.