Feature Clustering for Data Steering in Dynamic Data Driven Application Systems

Abstract

Abstract. In this paper, we describe how feature clustering on realworld cell-phone data can be used to locate the impact area of emergency events. We first examine the effect of two emergency events on the call activity in the areas surrounding the events. We then investigate how the time series of the affected areas behave relative to the time series of their respective neighboring areas. Finally, we examine the differences in hierarchical clusterings of the time series of the affected areas and neighboring areas. 1