Research And Application Of Ant Colony Clustering Algorithm

In the 21 st century of the information age,a large amount of data has been poured into different media.In many cases,it is difficult to directly obtain the hidden information behind these data.Cluster analysis is an important data mining method.It is involved in many fields such as economics,artificial intelligence,image recognition,and bioengineering.It has become one of the current research hotspots.The Ant Colony Clustering Algorithm(ACCA)is a bionic intelligent clustering algorithm.Compared with the traditional clustering method,this algorithm has the advantages such as excellent parallel computing,positive feedback,and self-organization,etc.,but it also has the following drawbacks: the convergence speed of the algorithm is slower at early searching stage,and the running time is longer;the transfer of ants is based on the number of pheromones,and the algorithm is prone to fall into a local optimum in the later period;the pheromone update lacks timeliness;The clustering quality is not ideal.This paper mainly proposes two kinds of improvements to the above insufficiency of ACCA algorithm.Aiming at the problem that ACCA algorithm is slow in early convergence and easy to fall into local optimum in the later period,an improved K-means ant colony clustering algorithm is proposed.Firstly,the K-means algorithm is used to preprocess the data set,and the distance matrix is quickly obtained.The obtained distance matrix is used as the basis for the ant colony clustering algorithm to initialize the pheromone matrix.Secondly,during the process of each ant iteration,a random perturbation variable is introduced into its clustering center,and several local search elements are generated as new clustering centers,then the data set is re-divided,and it is judged by the value of the objective function whether to retain the local search solution.Through the test of UCI data set,the results show that the hybrid algorithm does help to jump out of the local optimal solution.Aiming at the problem that the long running time of ACCA algorithm,the lack of timeliness of pheromone updating,and the problem of distinguishing the primary and secondary characteristics of each dimension of data,a two-stage ant colony clustering algorithm is proposed.In the first stage,the global pheromone update rule is combined with the real-time pheromone updates of each ant to strengthen the communication among ants,so that the ant colony can search for approximate solutions as soon as possible.In the second phase,membership matrix is used to adaptively assign different weight coefficients for each feature.The different weight coefficients distinguish the importance of each dimension and make the ant construct solutions effectively and improve the accuracy of clustering.After comparing the running time and accuracy of the artificial data set,iris,and wine datasets,the feasibility and effectiveness of the two-stage ant colony clustering algorithm are further verified.Finally,on the one hand,the two improved algorithms proposed in this paper are applied to the division of the economic strength of each province.The clustering results indicate that there are differences in the economic strengths of different provinces,which requires the country to implement effective economic policies and coordinate development;On the other hand,they are applied to image processing.The clustering and segmentation of simple digital image,gray character image and color brain image reflect that the improved algorithm in this paper has a certain practical value.