Research On Clustering Algorithm Based On Heuristic Intelligent Optimization

In recent years,with the rapid development of artificial intelligence technology,massive data has grown exponentially.As one of the important tools for mass data analysis,data mining technology has a broad application prospect.K-means clustering algorithm is the most common partition-based clustering algorithm in data mining.It has the advantages of simple and fast convergence speed.There are two problems in K-means clustering algorithm.The first problem is that the algorithm insensitivity to irregular clusters and unstable division of irregular clusters;The second problem is that the algorithm sensitive to the initial clustering center,and the clustering result is easy to fall into local optimization.Based on the above problems,this thesis proposes two improve algorithms,as follows:(1)In order to solve the problem of K-means clustering algorithm that is insensitive to irregular clusters and unstable clustering,the K-extended Central Point Clustering Algorithm is proposed in this thesis.First,the traditional maximum-minimum algorithm is improved,and the maximum distance is scaled down to select the center point,which reduces the probability of selecting outliers.Then,improve the measurement method.The basis for classifying data samples is changed to the proportion of distance,and the original hard distance is no longer used.Finally,the center point cluster is extended according to the discriminant formula.The experiment proves that the improved K-expanded central point clustering algorithm can not only ideally divide regular clusters,but also divide irregular clusters well.(2)In order to solve the problem of K-means clustering algorithm that is a strong dependence on the initial clustering center and cannot find the global optimal solution,the K-means Clustering Algorithm based on Bat Search Optimization is proposed in this thesis.First,introduce the sharing principle niche technology,increase the diversity of bats in the evolution process,introduce Levy flight random walk into the bat algorithm,speed up the search for local optimal solutions,optimize the optimization path in the convergence process,and thus improve the optimization ability of the algorithm accelerates the convergence speed.Multiple sets of simulation experiments show that the bat search algorithm based on the sharing principle niche technology is simple and efficient,and the convergence speed and optimization accuracy of the algorithm are improved while maintaining the advantages of the standard BA algorithm.Then the bat search algorithm based on the sharing principle niche technology and the K-means algorithm are fused to optimize the initial clustering center of the clustering algorithm.Experiments show that the algorithm has better stability and higher value in solving practical application problems.