| This thesis presents an attempt towards the improvement of the Standard Ant Clustering Algorithm by using the techniques of Genetic Algorithms. Goals of this thesis consist of multiple phases. The world of ants consists of two types of objects: artificial ants and data items. The task of the artificial ants is to wander around the world for a set number of steps, and attempt to form clusters for each type of data items. Next, ants pick-up a data item if it believes the location cell is not of a cluster. Additionally, if an ant is carrying a data item, it is expected to drop it off when it believes it falls within a cluster. During this process, any carrying ant (an artificial ant that is carrying a data item of any type) looks at a fixed neighborhood edge length to determine clusters existence. Edge length is anticipated to be relative to the world size, and it is not determined whether a larger or smaller edge would allow a higher clustering quality. This thesis will use the techniques of genetic algorithms and attempt to make use of biologically powered methods to maximize the clustering formation and come up with the best possible clusters that eventually will result into a new algorithm we will call ACAGA.. |
