Coverage constrained spatial Co-clustering

Given two geometric spaces, a set of matched points between two geometric spaces, the Coverage Constrained Spatial Co-clustering (CCSCO) problem produces k co-clusters that honor the coverage constraint and minimize the total distances of the spatial points to their cluster center. The CCSCO problem is important for many societal applications, such as design of evacuation routes and resource allocation. The problem is NP-hard; it is computationally challenging because of the large size of spatial points and the coverage constraint. This paper proposes a novel approach, called Bipartite Space Shrinking (BSS), for finding k clusters that minimize the total distances of the points to their cluster center under the coverage constraint. To improve the performance, we introduce the Distance Map data structure to efficiently construct a CCSCO. Experiments using real-world New York City Taxi Trip datasets demonstrate that the proposed algorithm significantly reduces the computational cost to create a CCSCO.