For efficient analysis of large smart city data sets in (real) time a knowledge-based approach is suggested, con- straining iteratively the search space while increasing accuracy by more complex computation. Results are shown for searching the distance to the nearest object, e.g., parking space.