This paper analyzes spatio-temporal interpolation methods based on shape functions, namely, the ST product and the tetrahedral methods. These methods yield data that can be represented and queried in constraint database systems. That is an advantage, because there are many constraint database queries that are not expressible in current geographic information systems. We illustrate and test our approach on an actual real estate database. The interpolations for house prices per square foot are compared on accuracy, storage requirement, error-proneness to time aggregation, and difficulty of representation. It is shown that the best method yields a spatio-temporal interpolation that estimates house prices (per square foot) with approximately 10 percent error, is less error-prone, and uses only linear constraints, which can be implemented in several recent constraint database systems.
展开▼
