Research On Affinity Propagation Clustering Algorithm Based On Manifold Distance And Density Adjustment

Due to the reduction of storage cost,Mass data has been emerging,the number of big database is more than the previous.Data mining technology has been more and more noticed in various fields.As one of the most important tasks of data mining,clustering technology can find potential information from the mass data without any priori information,which provides an effective basis for enterprise and government decision-making.Affinity propagation(AP)clustering as the most competitive clustering technology in the field of unsupervised learning has received broad attention in all fields since it was proposed.Even so,AP algorithm still has deficiencies.As affinity propagation(AP)clustering is sensitive to the complex structure dataset while calculating the similarity matrix and the cluster result is not ideal,an affinity propagation clustering algorithm based on manifold distance and density adjustment is proposed.The algorithm introduce local density of data and manifold theory into affinity propagation clustering,using a way of distance measure based on manifold structure and density adjustment to describe the clusters' actual structure better.So that it solves the similarity matrix's deficiency.At the same time,the algorithm is more efficient.Simulation experiment has been done by artificial datasets and standard datasets and the result shows the effectiveness and superiority of proposed algorithm.In addition,this paper constructs an evaluation model of agricultural drought disaster.A new classification model is built by combining the improved Affinity propagation with semi supervised idea.The model selects the 9 agricultural drought assessment indicators after the analysis of the research object.The agricultural drought data of Northern Anhui is the sample data.Compared with the BP neural network model,it is proved that this model can improve the accuracy of agricultural drought grade evaluation to a certain extent.