Pivoted Table Index for Querying Product-Property-Value Information

Abstract

The query for triple information on product–attribute (property)–value is one of the most frequent queries in e-commerce. In storing the triple (product–attribute–value) information, a vertical schema is effective for avoiding sparse data and schema evolution, while a conventional horizontal schema often shows better query performance, since the properties are queried as groups clustered by each product. Therefore, we propose two storage schemas: a vertical schema as a primary table structure for the triple information in RDBMS and a pivoted table index created from the basic vertical table as an additional index structure for accelerating query processing. The pivoted table index is beneficial to improving the performance of the frequent pattern query on the group properties associated with each product class.