SPARQL Query Optimization Based On Predicate Selectivity Estimation

The development and standardization of semantic technology have resulted in an unprecedented volume of RDF data being published,which led to the traditional data management techniques are no longer suitable for large RDF data.Therefore,the W3 C launched the standard query language SPARQL for RDF data.Faced with the large RDF data query processing and increasing complexity of query statements,improving the query efficiency of SPARQL has become a hot in the semantic Web.In practical applications,large number of SPARQL statements are with multiple predicates.Re-using the queries that with multi-predicates can reduces the number of repeated queries or search scope,which effectively improves the performance of SPARQL query.This paper proposes a SPARQL query optimization scheme based on predicate selectivity estimation.Based on the query logs,we construct a histogram to obtain the SPARQL predicate expression(SPE)statistics of history queries,then load cache according to the histogram.When a new SPARQL query statement is submitted to our system,it will first parse the query statement and get the corresponding SPE.If the SPE hits the histogram in whole or in part,system will directly return whole or part query result set from the cache to user,and the result set of remaining part that not hits the bucket will be fetched from underlying disk.The experiment demonstrates that the scheme based on query logs to build predicate histogram is effective in improving the query efficiency,thus achieving the goal of SPARQL query optimization.In addition,the experiment also confirms that the more predicates of the query statements,and the greater the amount of RDF data,the better of optimization degree.