Complex Network Clustering And Its Application In Neural Networks Research

Nowadays,with the rapid development of information technology, the amount of data is growing exponentially. How to find the information quickly and accurately from the large and chaotic data has become a very interesting topic. The clustering analysis method of data mining technology is one of the important methods, which provides a data research and analysis method for huge amounts of data and applies to various areas of real world widely. However, most of the cluster analysis algorithm requires pre-determined the number of clustering, how to determine the number of clustering is a complex and difficult problem.In recent years, artificial neural network due to high non-linear, parallel, good fault tolerance and strong robustness, etc., it has been widely used in many areas. Especially RBF neural networks have a strong ability of nonlinear fitting can map arbitrary, complex and nonlinear relation, but its ability of mapping nonlinear is reflected in the hidden basis function, the basis function of the characteristics of the main determined by the basic function of the center. However, the determination of basis function center is a difficult problem in practice.Based on the above issues, we discuss the clustering algorithm which combined the complex network community division techniques with the similarity measure in order to overcome the requirements of pre-determined number of clustering problem. Then introduce this algorithm into neural network to optimize the center value of the neural network. Finally, do the experimental verification.In this paper, we firstly discussed the advantages and disadvantages of the current complex network community division algorithm. Then we propose a clustering algorithm that combined CNM algorithm with similarity measure based on the research of k-means clustering algorithm. The algorithm this paper proposed overcomes the k-means algorithm's defect that needs to determine the number of clustering based on prior knowledge. Two cluster analysis experimental results show that the algorithm has improved the quality of clustering. Finally, we used this algorithm to optimize the center value of RBF neural network, two experiments prove that the algorithm overcomes the shortcomings of RBF neural network algorithm and improves the accuracy of the network.