Research On K-Means Clustering Method With Structural Constraint

With the rapid development of artificial intelligence and the rapid growth of data in the real world,it is of great research and application value to exploit the unlabeled data effectively.The K-means clustering algorithm is a classical algorithm for processing unlabeled data,which is a rather representative and widely used algorithm in the field of data mining.However,in the real world,the composition of data is often more complicated,and traditional clustering methods are difficult to adapt to the actual situation effectively and provide satisfying results for specific problems.In the unsupervised case,considering that the data in the real world are often associated with underlying structure,this paper focuses on how to effectively reveal or utilize the underlying structure of data for promising clustering.First,the K-means clustering algorithm is widely applied to image segmentation.However,traditional K-means clustering algorithms do not fully take into account the spatial constraint of image itself,or do not integrate spatial constraints into a unified clustering objective.To this end,we propose a general framework based on the K-means clustering method with image spatial structure constraints involved,considering the relationships of color space and spatial space jointly in clustering.In addition,existing data are often hierarchical in nature.The purpose of hierarchical clustering is to construct a hierarchical clustering tree that reflects the intrinsic hierarchical relationship between data.Existing methods are mostly designed heuristically without explicit objective function.For this reason,we propose a general framework with hierarchical spatial structure constraint for K-means clustering.When constructing the hierarchical tree,the clustering result of each layer can dynamically interact with each other,which is similar to the layer-by-layer updating of the neural networks.By studying the clustering method of K-means and considering the structure constraints of data,the proposed method outperforms existing methods on multiple data sets,which verifies the effectiveness of our method.By studying the clustering method of K-means and considering the spatial structure constraints of the data,the proposed method outperforms existing methods on multiple data sets,which verifies the effectiveness of our method.