ACO-based document clustering method

Ant systems are flexible to implement and give possibility to scale because they are based on multi agent cooperation. The aim of this publication is to show the universal character of that solution and potentiality in implementing it in wide areas of applications. The increase of demand for effective methods of large document collections management is a sufficient stimulus to place the research on the new application of ant based systems in the area of text document processing. Hitherto existing far generated ant based clustering methods are presented and briefly described at the beginning of that article. Next, the author defines the ACO (Ant Colony Optimization) metaheuristic, which was the basis of the method developed by him. Presentation of the details of the ant based documents clustering method is the main part of publication. 1. State of research on ant-based clustering methods Ant based algorithms are assigned to the group of multiagent systems. In such systems single agent (artificial ant) behavior is inspired by behavior of real ants. Ant based clustering and sorting algorithm Ant based clustering and sorting algorithm was first introduced by Deneubourg in 1990 [1]. As its name implies, two types of natural ant behavior are modeled by this algorithm. Firstly, clustering, where ants gather items to form heaps. An example for this is the clustering of dead corpses (cemetery formation) observed in the species of Pheidole pallidula. Secondly, sorting, where ants discriminate different kinds of items and spatially arrange them according to their properties. This type of activity can be observed in nests of Leptothorax unifasciatus, where larvae are arranged dependent on their sizes. In the Deneubourg’s model ants are modeled by simple agents, which randomly move in their environment, which is a square grid with periodic boundary conditions. Data items that are scattered within this environment can be picked E-mail address: [email protected] Pobrane z czasopisma Annales AIInformatica http://ai.annales.umcs.pl Data: 08/12/2017 07:22:39