Apply The Idea Of Density-Based Method To Antclass Clustering Algorithm

Data mining is the process of abstracting unaware, potential and useful information and knowledge from plentiful, incomplete, noisy, fuzzy and stochastic data, which is deemed to one of a foreland of data mining system and a promising cross-subject. Cluster analysis is one of the most important research domains of data mining. The motion of cluster analysis is that makes the data set into several clusters, a cluster is a collection of data objects that are similar to one another within the same cluster and dissimilar to the objects in other cluster. It has important appliances in may domain such as in business, biology, medicine, geography, web archive, and it is one of the hot research problem.In this paper, we have studied cluster analysis clearly, and apply the idea of density-based method to AntClass algorithm. The idea is that cluster is area having more points than other areas connecting with it. In this algorithm, the ants move objects in the 2D board frequently, by which the objects can be compared with each other. By this way we can place the similar objects together, remove the dissimilar objects away, and get the clusters that we are needed. Then we use K-Means algorithm using the initial partition provided by ants, which can converge faster. Because we consider the objects' attributes enough in the first time, the ants avoid moving unnecessarily during them move the objects. They only remove the objects in the low dense area and place them in the corresponding clusters, that is to say ants move the objects heuristically, which improve efficiency. This algorithm is also has random of ant colony algorithm, avoid clustering getting into local optimality, and avoid sensibility of the initial partition, for the initial input is got by the ant-based algorithm. It is the innovation of this paper that apply the idea of density-based method to AntClass algorithm.It can be found this algorithm is run faster than the ant-based algorithm by proving in theory and experimenting with some dataset, and the result is satisfmg.