Efficient Approximate Top-k Query Algorithm Using Cube Index

Exact top-k query processing has caught much attention recently because of its wide use in many research areas. Since missing the truly best answers is inherent and unavoidable due to the user’s subjective judgment, and the cost of processing exact top-k queries is highly expensive for datasets with huge volume, it is intriguing to answer approximate top-k query instead. In this paper, we define a novel kind of approximate top-k query, called approximation top-k query, and introduce an efficient indexing structure, cube index, to support this query. Based on cube index, we propose our novel algorithm: Cube Index Algorithm (CIA). We analyze the complexity of both setting up -cube index and CIA algorithm. Moreover, extensive experiments show that the CIA has significant improvement on the performance, compared with the well-known approximate top-k query algorithm, TA algorithm.