Output details
11 - Computer Science and Informatics
Goldsmiths' College
A probabilistic approach to mining mobile phone data sequences
<15> This paper presents a novel latent topic model developed as an extension of “Latent Dirichlet Allocation” addressing the incorporation of sequences, particularly long sequences representative of location data, into the topic model. The model has been evaluated on synthetic data as well as two types of real location data (GPS and cell tower connections). This paper was an invited journal paper, extending the nominated best conference paper at ISWC (IEEE Symposium on Wearable Computers 22% acceptance rate). Many research groups have requested the model code, including Telefonica and Southampton University. The paper provides an entirely new modelling strategy.