Research And Empirical Analysis On Community Detection Methods Of Technology Exchange Network

Scholars have successfully applied some of the ideas of complexity science to the network of description and interpretation of technology, but these attempts mainly remain at the macro level, and quantitative research on microcosmic and microcosmic aspects of technology trading networks is scarce. In contrast, the network structure of the community in-depth and meticulous analysis, can be more comprehensive and profound understanding of the structure of the system itself, the law is conducive to building a good technology trading patterns, and promote the smooth conduct of technical transactions.In this paper, the community detection algorithm is applied to research the process of technology exchange behavior, and we use the actual transaction data to fit its' generation model, and have empirical analysis on the community detection algorithm through the community structure of the real technology exchange network. The main content of this dissertation is summarized as follows:(1) Construct the technology exchange network model. After constructing the complex technology exchange network, we can have a preliminary understood of the network structure from the static characteristic and the dynamic characteristic. Select the most appropriate community detection method to discover the laws between the evolutions of the community structure and the transaction activity according to the time dimension.(2) Do research and empirical analysis on the community detection methods of technology exchange network. We compare four kinds of community detection algorithms, which based on hierarchical clustering,modularity optimization, spectral clustering and streaming. We find modularity optimization algorithms are good at modularity and time complexity after compared with other four algorithms. For the technical transaction network, the cohesion coefficient is higher and the structure tends to be stable through the time dimension.(3) Pick the important nodes and communities of technology exchange network to find rules. A new centrality evaluation parameter is obtained by weighting various centralities. Through the evolution of the important nodes, it is found that the internal nodes attract the external nodes into the community mainly through the important nodes, but the direct links between the important nodes are sparse. Through the cluster of the network within the community, the node degree of large groups out more branches, nodes of small associations are more closely linked.In this paper, we use the modularity optimization algorithm to classify the communities based on the Beijing technology exchange network. We find that the algorithm has better performance in terms of time complexity and modularity, and the internal structure of the community found by algorithm can provide guidance and support to the management of technology trading market to a certain extent.