Rough Clustering Algorithm Based on Data Field
Qinrong Feng · 2009
Clustering analysis is the hotspot in Data mining,all the conventional clustering algorithms precisely put the each object into one cluster,the bounders between clusters are precise,as the development of the Web mining,clustering algorithms that precisely divide each object face great challenges.Based on the data field theory and classic rough set theory's character that processes the uncertainty and imprecise data,a novel rough clustering algorithm based on data field was proposed,it divides the objects through computing potential value,which avoids the conventional rough clustering partition method based on euclidean distance.The approach iterates from rough to un-rough incessantly till the stable clusters form.At the experimental analysis process,we compared the algorithm that we proposed with rough K-means algorithm and rough K-medoids algorithm,the result shows the algorithm that we proposed has better clusters on the crossed datasets and fast convergence.