SimMatching: adaptable road network matching for efficient and scalable spatial data integration.

Spatial data integration is a challenging task due to the high degree of diversity between different geodata sources, the inherent complexity of objects, and the large size of datasets. To avoid duplicates in an integrated dataset, input sources have to be linked on the instance level. By matching spatial objects, multiple representations of the same real-world entity shall be identified based on similarity computation. In this paper, we present an approach for similarity-based spatial matching of road networks. Our SimMatching algorithm adapts to a variety of input data characteristics by using weighted similarity measures. Geometric and semantic attributes are considered as well as the dataset topology to enhance similarity computations with relational measures. We use a greedy approach and optimizations to keep the number of match candidates minimal all the time. This allows very low runtimes while giving high quality matching results. Supported by a partitioning framework and parallel processing, it also guarantees scalability to large datasets.