Study On Clustering Algorithm In Data Mining

Facing the massive volume and high dimensional data how to build effective and scalable clustering algorithm for data mining is one of research directions of data mining. Aiming at above issues, some clustering algorithms have been studied substantially as follows:A Tree-structured Growing Self-Organizing Map (TGSOM) model is presented as an extended version of the Self-Organizing Feature Map (SOFM), which has a dynamic structure generated during the training process. Experiments show that this model can achieve hierarchical clustering of a data set, and require fewer nodes to represent the data set and less processing time compared with SOFM and Growing Self-Organizing Map (GSOM).A clustering algorithm (CUBN) is presented, which integrates density-based, gird-based and distance-based clustering methods. The experimental results show that CUBN can identify clusters having non-spherical shapes and wide variances in size, and its computational complexity is linear-time, so the algorithm facilitates the clustering of a very large data set.A clustering algorithm (CMM) is presented, which is more robust to outliers, and can identify clusters having non-spherical shapes and wide variances in size. CMM achieves those by representing each cluster using multiple medoids. CMM is also a linear-time clustering algorithm, and therefore, it facilitates the clustering of a very large data set.A clustering algorithm (RDVS) is presented, which integrates reference, density, and neural network methods. It captures the data set's character by references, and then input the position and density of references into the neural network. RDVS keeps the clustering ability of reference method, and can get a better visual clustering result by using Visualization Self-Organizing Map (ViSOM).