| We study the optimal-location queries in spatial databases. Given a set S of sites, a set O of objects, and a spatial region Q, the optimal-location query returns a location in Q such that, if a new site is put at that location, the total benefit to the objects is maximized. The optimal-location queries are of interest in many applications such as corporation decision-support systems. There may be many variations of the optimal location because different applications may define the optimality differently. In this thesis, we consider the geometric proximity between objects and sites and study two most intuitive definitions of the optimality, namely Max-Inf and Min-Dist. Max-Inf measures the total benefit as the number of objects closer to the new site than to any existing sites and aims to maximize it. Min-Dist measures the benefit as the savings of the distance from objects to their nearest sites and aims to minimize the average distance from each object to its nearest site. We also examine the problem under three different distance metrics, namely L1, L2 and network shortest path. We propose efficient solutions for each optimal location query (6 in total) and evaluate them experimentally. We expect that the work will not only advance spatial database research, but also benefit the end users who deal with a large volume of spatial data. |
