Improved Affinity Propagation Clustering Algorithm Based On Multiple Theories And Its Applications

With the development of society and the wide spread of Internet technology,the demand for related data in various industries is increasing steadily.In fact,the emergence of big data era is the product of the perfect combination of the growing social demand and the matching data mining technology under the background of the rapid development of Internet technology.Clustering analysis is an important method in the field of data mining.The affinity propagation clustering algorithm(AP),as a method that can efficiently and quickly complete data clustering analysis in the field of data mining,has been widely used to research and promote by experts and scholars at home and abroad.However,the traditional affinity propagation clustering algorithm still has many disadvantages,resulting in low clustering performance,poor clustering effect and so on.Based on this,this paper introduces a variety of different theories into the original nearest neighbor propagation clustering algorithm,and tries to obtain better clustering results to improve the clustering performance of the algorithm.The main contents of this paper are as follows:(1)Taking the similarity calculation method as the starting point,aiming at the disadvantages of the traditional AP algorithm which is sensitive to the data type,this paper proposed a novel self-adaptive affinity propagation clustering algorithm based on density peak theory and weighted similarity(DPWS-SAAP).In the traditional AP algorithm,the local density theory of the density peak clustering algorithm is introduced,and the density attribute is constructed in the original algorithm.The similarity is calculated by using the idea of weighted similarity,and the similarity matrix is updated finally.(2)According to the traditional AP algorithm,Euclidean distance is the only way to calculate the similarity of the algorithm,which can not better reflect the characteristics of the data sample space.At the same time,it is impossible to get accurate clustering results.Gravity theory-based affinity propagation clustering algorithm(G-AP)is proposed.This algorithm introduces the idea of gravitation,and further optimizes the similarity calculation of the algorithm and formula to get a better clustering effect.(3)As a new and efficient swarm intelligence algorithm,firefly algorithm is widely used in many fields.In this paper,the affinity propagation clustering algorithm based on firefly optimization(FO-AP)is proposed,which optimizes the adaptive scanning preference space and updates the similarity matrix to obtain the optimal clustering.(4)In view of the disadvantages of traditional AP algorithm in processing high-dimensional data,it is difficult to obtain effective clustering results.This paper proposes affinity propagation clustering algorithm based on improvement distance function(IDF-AP)to improve clustering performance.And the accurate clustering is obtained.