Seminar by Dr Damien Fay: "GlasDasha: predicting arrival times from crowd sourced Smartphone data without localisation or route maps"

Our speaker is Dr Damien Fay, Lecturer at BU, former mathematics Lecturer at the National University of Ireland and senior research associate in Cambridge and Cork.

Dr Fay has a very good track record in areas such as time series prediction using a wide range of approaches, Gaussian processes and applied graph theory to cite a few.

Abstract: Accurately predicting the arrival of occupants to a building would enable more efficient HVAC control and improved occupant comfort. We present an approach to arrival prediction through the use of Smartphones and WiFi access points. The techniques use the concept of a field, representing expected travel times from access points to an anchor location for different models of travel. The field is constructed from crowd-sourced data. Users maintain privacy by not revealing location information, but instead using offline reporting of access point IDs. The system is light on battery use, handles multiple travel modes, provides tailored predictions for individuals, and identifies deviations from normal behaviour.
We demonstrate the effectiveness of the approach by a live field study using data collected over a period of five months. There is also a large data analysis section to this talk which includes extracting modes from distributions using Gaussian Mixture Models, smoothing and in addition outlier detection.