Stock Clustering Based On Complex Network

The stock market is a very large complex system,the domestic stock market is more complex,trading volume and price instability.In recent years,there are more and more cross-shareholdings and mutual investment cooperation in the stock market,which makes it more difficult to analyze stocks.Clustering stocks and discovering the relationship between stocks will help investors make more scientific and reliable investment decisions.Spectral clustering algorithm is based on the theory of complex networks and its core is to transform the clustering problem into the optimization problem of graphs.Compared with the traditional clustering method,spectral clustering can be applied to data of any shape and converges to the global optimum.However,one disadvantage of this algorithm is that it is sensitive to scale parameters.This paper combines ensemble learning and spectral clustering,which can not only overcome the problem of selecting scale parameters of spectral clustering,but also construct a variety of base learners,making the clustering results more stable and reliable.In this paper,the complex network is applied to analyze the fluctuation characteristics of the sse50 stock price,eliminate the market factors,construct the similarity matrix through the network autoregressive model,and establish the complex network diagram.With the spectral clustering algorithm as the learning machine,the voting method is skillfully integrated.From the clustering results,we also find some inherent relations that cannot be found by traditional clustering algorithms.