Bayesian Prototype-Based Clustering for Spatio-Temporal Data

• Activity Number: 136
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Angela Schoergendorfer*+ and Huijing Jiang
• Companies: IBM T. J. Watson Research Center and IBM T. J. Watson Research Center
• Keywords: spatio-temporal data ; clustering with fixed prototypes ; hierarchical model ; functional data analysis ; data center management

Abstract

A new Bayesian method is introduced for clustering spatial locations based on measurements taken over time when a fixed set of prototypes or underlying patterns is given. The proposed procedure employs a Bayesian hierarchical modeling approach incorporating spatial dependence and prototype information explicitly. Moreover, a functional data analysis approach is utilized to model patterns over time, and to accommodate asynchronous/missing data. This explicit model-based method can provide a probabilistic cluster membership for any location at which no measurements are available. The proposed clustering procedure is illustrated using a data center management problem in which thermal zone maps are developed and updated dynamically via real-time sensor information to monitor the influence of each cooling source on the entire data center.