On Effectiveness of Database Accessing Methods for Subset Searching

The field of relational database research has developed many database accessing methods for effective data retrieval, e.g. B tree indexing, Bitmap indexing, hash-based join, sort-merge join. These methods are oriented on finding or joining single items (represented by records) that satisfy point or range conditions. However, in the area of data mining research, there is often the need to effectively retrieve the multi-item sets that contain a given multi-item subset. We will refer to this type of retrieval as Subset Search Problem. In this paper we formally define the Subset Search Problem. We analyze and experimentally verify the usefulness and effectiveness of the most common database accessing methods applied to the Subset Search Problem. The results show that Bitmap indexing gives the best improvement to the implementation of the Subset Search Problem.