Data mining of home data revealing lifestyle changes

The goal of telecare is to augment caregivers in a future where human resources will be under strain. The ‘Dolls House’ project at Dundee University explores the concept of sensor-equipped homes through a custom-built model equipped with standard sensors. Generating realistic volumes of data is time consuming, yet vital in this research area.

SynSensor was developed to provide a solution to the generation of realistic sensor data through a platform on which personas can be created to emulate months of sensor data with natural perturbations and trends. SynSensor generates large quantities of behavioural data that can be manipulated through modulation algorithms to synthesise a realistic noise floor and to introduce behavioural trends into the data.

Given that behavioural changes can result from associated changes in health, data mining tools could provide insight into an individual’s ‘well being’. Test data, containing a known behavioural trend, was analysed with standard data mining tools to determine if they could detect the presence of the trend. Application of SQL Server’s data mining tools proved inconclusive and suitable data aggregation and segregation methodologies are discussed to aid further research. SynSensor achieved its goal of generating synthetic data in volume, facilitating future study into this area.

Christopher Walker.