Clustering Algorithm Based On Ant Colony Algorithm

Through modeling the ant can always find the shortest path from their nest and food by mathematical method, we can obtain a new algorithms– ant colony algorithm. This algorithm has several virtues such as positive feedback and parallel mechanism, preferable robusticity, distributed computing, and easy combination with other methods. And presents outstanding performance and huge development potential in solve complex problems.With increasing use of computers, tremendous volumes of data have filled hard disks as digitized information. In the presence of the huge amount of data, the challenge is how to truly understand, integrate, and apply various methods to discover and utilized knowledge from these data. To predict future trends and to make better decisions in science, industry, and markets, people are starved for discovery of knowledge from this morass of data. Clustering analysis is one of the most important theories in data mining. The goal of clustering analysis is to group similar objects together. The grouping is done such that patterns within a cluster are more similar to each other than patterns belonging to different groups.First, we are review and introduce the theory of clustering as detail as possible. Analysis and summary the strengths and weaknesses of clustering algorithm. And also introduce data type and measurement of clustering configuration. Then we introduce the basic ant colony algorithm and summary the two model application in clustering problem. The main contribution of this paper is proposed a new ant colony based clustering algorithm. The algorithm considers R ants to build solutions. An ant starts with an empty solution string S of length N where each element of string corresponds to one of the test samples. The value assigned to an element of solution string S represents the cluster number to which the test sample is assigned in S . In order to escape the local minimum solution, we employ local search procedures. Thus, the algorithm repeatedly running for a maximum number of given iterations, and solution having lowest function value represents the optimal partitioning of objects of a given dataset into several groups. In order to prove the effectiveness and superiority of this proposed algorithm, we employ genetic algorithm, tabu search approach, simulated annealing approach running the same dataset which are running by proposed algorithm. The dataset include 2 simulated dataset and 3 dataset from UCI. The result indicate the proposed algorithm is more effectiveness than the other 3 algorithm.As the information on website is more and more abundant and the toplogy of it's more and more complex,"information overloading"and"resource maze"in information service are ubiquitous. Different users have different accessing intention. Therefore, the application of the ant clustering algorithm to web usage mining can help to serve more efficiently through implementing adaptive websites, namely, transforming the knowledge extracted from Web usage data into website intelligence. Thus , we design and realize a prototype system of adaptive website based on web mining technology.At last, we conclude and analyze current research work, discuss our intending research work in this area.