Unsupervised Clustering of Context Data and Learning User requirements for a Mobile Device

J. A. Flanagan. Unsupervised clustering of context data and learning user requirements for a mobile device. In A. Day, B. Kokinov, D. Leake, and R. Turner, editors, Modelling and Using Context: 5th International and Interdisciplinary Conference CONTEXT 2005, Paris, France, July, Proceedings, volume 3554 of LNAI, pages 155–168. Springer-Verlag, Berlin Heidelberg, 2005. [url]


This paper present a technique for unsupervised learning algorithm for inferring important locations to be associated to a posssible custimization of the user interface of an application. The author departes from previous studies of inferring context extending the technique in such a way that would be possible to use the method on-the-fly and without the training period, which is impeding the development of such techniques on mobile devices.

The algorithm is called K-SCM and it is based on the idea to fuse several input sources into a string. The string is then associated to a matrix for which is calculated the variation and probability of presentation of a certain node at a certain time. Using solutions from the neural networks theory, the algorithm is able to select the winning node over a certain period of time that will define the winning location to be associated with the user interaction on the phone.


Leave a Reply