Selectivity estimation for SPARQL graph pattern

This paper focuses on selectivity estimation for SPARQL graph patterns, which is crucial to RDF query optimization. The previous work takes the join uniformity assumption, which would lead to high inaccurate estimation in the cases where properties in SPARQL graph patterns are correlated. We take into account the dependencies among properties in SPARQL graph patterns and propose a more accurate estimation model. We first focus on two common SPARQL graph patterns (star and chain patterns) and propose to use Bayesian network and chain histogram for estimating the selectivityof them. Then, for an arbitrary composite SPARQL graph pattern, we maximally combines the results of the star and chain patterns we have precomputed. The experiments show that our method outperforms existing approaches in accuracy.