Complex Dynamic Meta-Network Model And Its Applications In International Trade

In the academic research field of International Trade Network, most researchers use macro-level data to establish network model, i.e., the country as a node, trade relations of import or export between countries as arcs; and then used the static graph-based measures to explained the patters of trade flows among countries, as well as the issue on economic globalization. In the process of international trade, the internationalization behavior of enterprises will affect the international trade network of a country or a region, such as, the emergence of networks, network changes over time, and so on. Therefore, it is necessary to use micro-enterprise level data to study the international trade network. This paper proposed an interdisciplinary approach to study international trade networks, by using enterprise-level data modeling and analysis.However, in the micro-level of international trade activities, because of involving a variety of elements and complex relations among different elements, for example, enterprises, technology, products, markets and trade organizations, etc.. Moreover, the system of international trade exhibited characteristics of complex systems which is large-scale, multiple-elements, and multiple-mode and multiple-plex relation. Traditional one mode or two mode network model cannot support complex mode network modeling. Therefore, to meet the needs of micro-level international trade network modeling, this paper first proposed a international trade complex metanetwork model by extending Charley’s metanetwork model to the complex metanetwork model, and then proposed a dynamic network framework, named the International Trade Complex Dynamic Metanetwork Model, to describe the dynamic nature of the overall metanetwork and the factors caused nodes and links change.As the application of international trade complex dynamic metanetwork model need to handle multiple data sources and massive data streams, and facing challenges of complex data types, this paper proposed an innovative modeling solution: establish an Business Intelligent system which composites with database, data warehousing, OLAP and data mining, in order to support complex metanetwork modeling and analysis.Finally, based on high-tech export trade data of Guangdong province during the period of 1995-2004, this paper established the complex dynamic network composing of ten metanetworks, and then used time-series network analysis, multiple variables statistical analysis and data mining: (I) analyze temporal changes of the whole metanetwork topology properties, the structure evolution of sub-metanetwork or subnetwork; (ii) detect the changes of the network topology prosperities or the network structure; (iii) mining the enterprise node’s behavior patterns; (iv) evaluate and classify enterprise nodes based on graph measures; (v) indentify market association structure hidden in the relation data between enterprise and export markets.The innovation of the present paper is chiefly embodied in the following aspects: (i) it proposed a complex metanetwork model for international trade network modeling by using micro-level data coming from enterprise, which offers a new model for the research of international trade network. (ii) It proposed a methodology to support the large-scale complex dynamic metanetwork modeling by using powerful techniques provided by Business Intelligent platform, which will provide a methodological reference for the similar research. (iii) It proposed a dynamic bipartite network representation schema by enhancing edge definition with timestamp, considering the bipartite network is the“gene”network in complex dynamic metanetwork. And then it applied data mining techniques to discover the bipartite subgraph, the linkage pattern and the nodes behaviors. (iv) It studied systematically the high-tech export trade of Guangdong based on the above works. The empirical results provided large information on the structure prosperities of export trade system. These rich information cannot be provided and explained by the traditional economic statistical methods.