Research On Clustering Algorithm Based On The Fuse Of Genetic-ant Colony Algorithm

Computer has been applied so widely to practical production that there have been massive data,and the door of the era of big data is opened.Thus,there is a challenge that how to uncover the valuable and complete knowledge that provide an objective basis for decision making,and this challenge is making data mining technology grow even faster.Cluster analysis is one of important technique in data mining and it has been widely used in many fields,such as data analysis,pattern recognition and image processing.In recent years,ant colony algorithm has been applied in cluster analysis.Ant colony clustering algorithms has positive feedback and high parallel characteristics,but some aspects are discussed further by us,such as slow convergence,prone to stagnation and easily gets into local optimal structure.However,genetic algorithm could improve the global search ability of the algorithm rapidly and effectively.In this thesis,the researcher integrate these two algorithms and make uses of the advantages of two algorithms in order to have more effective way to carrying out cluster analysis.The main results are as follows:First of all,the dissertation introduces knowledge of cluster analysis,and study the Ant Colony Algorithm and Ant Colony Clustering Algorithm related knowledge.On the basis of that,GACA algorithm have put forward.Secondly,based on the combination of Genetic Algorithm and Ant Colony Algorithm,a novel design idea has been presented,which is in terms of single element and heap form distributes in Two-dimensional space.In addition,ant memory marker strategy is used to prevent the ant transport data occurs shaking repeatedly phenomenon.In the Genetic Algorithm,the researcher set up max and minimum evolution algebra by population evolution rate determine algorithm fusion point,and attempt to a research idea that uses clustering quality to estimate clustering number.Finally,this dissertation presents four UCI data sets which are Car,Soybean,Voting and Zoo,and experiments have been carried out using K-mode,ECA of clustering algorithm based on information entropy,GACA and LF four kinds of algorithms.Based on the research to the results of the experiment,it follows that GACA algorithm has better accuracy and shrinkage in clusters aspects.