A Tabu Search Approach to Clustering

Abstract

Summary. In Clustering Problems, groups of similar subjects are to be retrieved from large data sets. Meta-heuristics are often used to obtain high quality solutions within reasonable time limits. Tabu search has proved to be a successful methodology for solving optimization problems, but applications to clustering problems are rare. In this paper, we construct a tabu search approach and compare it to the existing k-means and simulated annealing approaches. We find that tabu search returns solutions of very high quality for various types of cluster instances.

Blending multiple algorithmic granular components: a recipe for clustering

“…An example of a move is reassigning the cluster membership of a data object in each of the k clusters that make up C c t . The reassignment of cluster membership can be based on probability [Al-Sultan and Fedjki, 1997] or deterministic strategies [Turkensteen and Andersen, 2009]. A subset N * (C c t ) of the neighborhood set is then selected -a basic selection mechanism is finding the difference between the neighborhood set and the tabu list…”

“…, c k(0) } consisting of clusters having data objects from S assigned to them. The initial solution can be randomly generated by randomly assigning each data objects to exactly one cluster; or obtained using formal means such as the Ward's algorithm in[Turkensteen and Andersen, 2009]. The current solution C c t and the best solution C b t are then set to the initial solution C c t = C 0 , C b t = C 0 (the current solution C c t is the solution in context at iteration t; the best solution C b t holds the overall best solution at any given instance and eventually becomes the global optimal solution).…”

K-Means Cloning: Adaptive Spherical K-Means Clustering

“…Such algorithms exploit guided random search. Therefore, much better results have been obtained with meta-heuristics approaches, such as simulated annealing [9][10][11][12][13][14], tabu search [15,16], and genetic algorithms [17]. They combine techniques of local search and general strategies to escape from local optima leading to broadly searching the solution space.…”