Reserach And Application On The Clustering Algorithm

Data mining is a process that in the vast amounts of data looking for patterns or rules.Data clustering is an important data mining technology for people to understand and explore the inherent relationship between things.It can be used as independent data mining tools to get some in-depth information from the knowledge base, but also can be used as the pretreatment step for other data mining algorithms.It is widely used in business management, market analysis, engineering design and scientific exploration and other fields.Clustering is to partition data objects into different categories, or clusters, making the similarity with the clusters of data as small as possible, while the dissimilarity of different clusters of data as large as possible.Ant colony algorithm and particle swarm optimization are the two main algorithms in the field of swarm intelligence.Ant colony algorithm originates from the study on the accumulation of their bodies and classification of their larvae.Particle swarm optimization originates from the study on the group movement behavior of birds.As a new evolutionary computation technique,swarm intelligence algorithm has become more and more researchers focus of attention, and was introduced to the field of data clustering and plays a huge role.However, ant colony algorithm and particle swarm optimization is still not perfect,there exists their own intelligence deficiencies in the process of data clustering, resulting in unsatisfactory clustering effect.Thus how to design effective clustering algorithm will be an important topic in the field of swarm intelligence theory in the development of clustering. Aimed to solve the problem that the ant-based clustering algorithm expenses a long time and easily produces redundant number of clusters,a new multi-loading ant colony clustering algorithm based on dynamic neighborhood is provided. The algorithm seeks for the pure neighborhood through the neighborhood dynamic auto-adapted adjustment, enhances ant's memory to store the size of the pure neighborhood, exchanges the information between ants, multi-loads and merges the similar regions to form the final cluster result.Experiment shows that the new algorithm effectively advances the efficiency of algorithm and the result of clustering.In view of the advantages and disadvantages of K-harmonic means and chaos particle swarm optimization clustering algorithm,a chaos particle swarm optimization clustering algorithm merged with K-harmonic means is provided.Particle swarm was firstly divided into sub-groups, resulting in an iterative process for each particle based on its extreme value and location of the individual sub-populations in the global extremum to update their position. Secondly, the introduction of variable-scale chaotic mutation, inhibit the premature convergence and improve the calculation accuracy. The new algorithm can avoid the algorithm into a local optimum, guarantee the convergence speed while enhancing the capacity of global search algorithms , improve the clustering efficiency.