This technical report describes the homeset algorithm, a simple yet effective 
approach to estimate home occupancy schedules from unlabelled sensor data. The 
algorithm relies on Wi-Fi scan data to determine when residents are at home 
and when not. We validate our approach using a data set from the Nokia 
Lausanne Data Collection Campaign that contains mobile phone traces of 38 
participants collected over more than one year. Since the data is unlabelled, 
we indirectly validate our results leveraging the information hidden in 
anonymised GPS traces collected by the mobile phones of home occupants. We 
further show that the homeset algorithm is able to autonomously determine the 
reliability of the computed schedules. Finally, we show how these schedules 
can be used to predict the future occupancy behaviour of mobile phone owners.