Application Of Improved Ant Algorithm In Spatial Data Clustering
Hazhong Qian,Fang Wu,Lei Ge,Bo Chen,Huilian Wang
2006-01-01
Abstract:Spatial clustering is important application of artificial intelligence in spatial information system. A new clustering method based on grid technique and improved multi-variety ant algorithm was developed, which was expected to satisfy the requirement of clustering speed, intelligence, and independence in spatial information system. First, the clustering's basic concepts and developments were introduced. The shortcomings of existent clustering algorithms were analyzed, and the steps of improving these algorithms were expatiated. Second, a way of transforming vector data to grid was presented, and a formula of calculating the value of grid unit was defined. Third, according to the ant's character of multi-variety, three type ants and their mutual-transformation modes were defined, which were used to regulate ants' action and improve algorithm's intelligence and agility. Fourth, based on the grid technique and multi-variety ant algorithm, a new clustering method came into being, and its whole process was expounded. At the end of this article, an example was illustrated for clarifying the theory of this method. Analysis indicates that this method solves the main problems in existed algorithms, and is more ascendant for spacial data clustering.