Discovering Patterns of Insurgency via Spatio-Temporal Data Mining
Abstract
The need to discover patterns in spatio-temporal (ST) data has driven much recent research in ST cooccurrence patterns. Early work focused on discovering spatial patterns such as co-location without examining the development of patterns over time or the temporal aspect of ST datasets. This paper describes a novel set of cooccurrence patterns called mixed-drove co-occurrence patterns (MDCOPs). They represent subsets of two or more different ST object-types whose instances are close to each other both spatially and temporally. However, mining MDCOPs is computationally very expensive due to complex interest measures, larger archived and historical datasets, and exponential growth in candidate patterns with the number of object-types. We propose a monotonic composite interest measure or discovering MDCOPs and two novel MDCOP mining algorithms. Analytical results show that the proposed algorithms are correct and complete. Experimental results also show that the proposed methods are computationally more efficient than naive alternatives.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2008
- Accession Number
- ADA505835
Entities
People
- James A. Shine
- James P. Rogers
- Mete Celik
- Shashi Shekhar