Bi-tour ant colony optimization for diagonal clustering

Abstract

Clustering data represented by two-dimensional matrix can be achieved by re-arranging the permutation of rows and columns into diagonal block structure. The non-overlapping fragments are then determined by identifying the blocks formed in the revised matrix. In this paper, we propose using Bi- tour Ant Colony Optimization (BTACO) for diagonal clustering, which is a variant from the Ant Colony Optimization algorithm. In the computational study, we show that BTACO can re-arrange the matrix into diagonal block structure well for vertical partitioning in the relational database system.