Clustering Algorithm Research Based On Self-Organizing Feature Map Network

Data mining is the untrivial procedure of extracting of implicit, unknown, potential useful knowledge and patterns that are not explicitly represented in the database, which is widely applied in GIS, remote sensing, image processing, navigation system and many other fields in recent years. However, extracting knowledge from databases is more difficult due to the huge amount of data and the complexity of data type and relationships. A crucial challenge in data mining is the performance and efficiency of data mining algorithm, the improvement of which is the key to data mining techniques'advancement, innovation and application.Clustering is the task of grouping the objects of a database into meaningful subclasses (that is, clusters) so that the members of a cluster are as similar as possible whereas the members of different clusters differ as much as possible from each other. Due to its unsupervised learning ability, clustering has been widely used in numerous applications, such as pattern recognition, image processing, market research and so on. Clustering can find out the dense or sparse areas of data distribution, which can help to discover the distribution mode and interesting relationship from data.However there exit many limitations of the exiting clustering algorithms, such as not highly scalable, failing to discover clusters with arbitrary shape, sensitive to the order of input data and can not deal with multi-dimensional data, which will be further researched in the paper.Firstly, the main clustering algorithms exiting are analysed and researched in the paper. And simulated experiments are carried out to discuss performance change according to important parameters. The advantages and disadvantages of the algorithms are summarized.Secondly, self-organizing map network are researched and the characteristics of the netowork are figured out through simulated experiments.And then, a novel simility measure called topology similarity is proposed in the paper. A clustering model is built based on this simility measure, which take good use of the concept, topology similarity matrix.This paper proposes a novel clustering algorithm based on a family of self-organizing feature map network according to the features of self-organizing feature map network. The algorithm is presented in detail in the paper. The simulated...