Top-k Search Over Grid File

In the era of huge datasets, the topsearch becomes an effective way to decrease the search time of topobjects. Since we suppose locally accessible data only, the multidimensional indexes containing all attributes together seem to be more effective than a distribution of each attribute to a separate index. Therefore we introduce the topsearch algorithm over grid file – the multidimensional index not used for the topsearch yet. Grid file does not require computation with all attribute values together like R-tree, R*-tree (i.e. computation of area, perimeter) nor a metric like M-tree. Grid file can be used directly for indexing any type of attributes with natural ordering. Our experiments show that grid file, R-tree and R*-tree offer much better performance of the topsearch than separated B-trees and table scan.