The Study Of Boundary Detecting Algorithm Based On A2-MST And Ensemble Boundary

Clustering is one of the most important technology in data mining and a central issue in academic research,and play an significant role in various fields of data analysis.The clustering boundary identification can improve the accuracy of clustering results and reveal the the characteristics of clustering itself. Accordingly,the research about that become an hot spot increasingly,which has already exploited in market analysis, fraud detection, customer retention, product control and scientific exploration urgently.The current clustering boundary detection algorithms are always sensitive to parameters,inefficient implementation and can not extend to high-dimensional space easily. Contrary to these problems,this essay will do a deep research in clustering boundary detection algorithm.In order to improve the efficient of clustering boundary detection algorithm,we proposed a two road spanning tree boundary detection algorithm based on two road spanning tree technology cluster datasets,then proposed a new concept of hierarchical neighbors which means to construct a similarity graph on the two road spanning tree,therefore the hierarchical neighbors of a node was a sum of the data nodes number which located around the few layers of the data node.Noticed that the number of layer is installed by the user.Finally,according to the number of hierarchical neighbor judge the boundary points.Experiments proved the efficiency of this algorithm’s the implementation,it was able to identify the boundary of arbitrary shape, and has a double effects in clustering and boundary detection.Unlike other boundary detection algorithms, clustering and boundary detection algorithm was separated.Ensemble boundary detection was proposed because of existing clustering algorithms sensitive to parameters and with low accuracy. Similar to ensemble clustering, in order to avoid non-supervised learning to make a mistaken assumption without any the prior knowledge,ensemble boundary ensembles the existing boundary detection algorithms to extract the finally boundary results,which means to get a correct boundary results from integrating multi-boundary result based on similarity graph and consensus function. Another advantage of his algorithm is that user does not enter any parameters.The experimental results demonstrate that the algorithm can effectively handle multiple density, arbitrary shape and size of dataset.