Mining Of The Hidden Capability Network For Product Space In The Context Of Big Data

Study through complex networks on big data which accumulated of different disciplines has become a hot topic of current research. According to the problem where national economic development potential from, Hausmann and Hidalgo et al put forward the theory of comparative advantage evolution which from product space by analyzing the similarity of nation-product export network based on the data of international trade, and points out that the evolution direction of comparative advantage and degree of development of a country is determined by the location of its product space in the overall product space of world. In the analysis of reasons why backward countries are difficult to achieve leapfrog development to catch up developed countries, they proposed the concept of competency model, and consider that the state and the products are both collection of capabilities. And they constructed the random three-dimensional network of state, products and the capabilities(hereinafter referred to as the hidden network), demonstrated nations of less capability will be difficult to accumulate more capability while nations of more capability will increase its capability easily amount through the accumulation of new capability then obtain higher returns. But they did not find a method to mining the hidden network based on empirical data, so that they have some difficult to explain the current situation existing in the real world.In this paper we consider that the process of mining hidden capability network is link prediction, and find a way to construct the hidden capability network through the genetic algorithm based on the concept of capability model, and we certificate the rationality of nation-capability network and product-capability network through analyzing the capability model. Finally, based on the capability model we calculate the index of potential of nation’s economic complexity.The main work includes the following aspects:(1)Introduce the link prediction theory into the mining of the hidden capability network.This paper analyzes the advantages and disadvantages of the traditional link prediction methods for mining of the hidden capability network, and put forward the idea that we can introduce the general framework of genetic algorithm, using link prediction specific means for mining of the hidden capability network.(2)Implementation and improvement of genetic algorithmThis is the core of our work. Due to the random search space of constructing a reasonable hidden capability network is too large, to reduce the searching time, a number of a priori knowledge is introduced in the definition of objective function and algorithm sequence, then we received a satisfaction result.(3) The rationality analysis and application of the hidden capability networkAs to judge hidden capability network construction is reasonable or not, we analyzes the rationality for search results, we also test it by the conclusion obtained by Hausmann and Hidalgo based on binomial random network model about universality of capability model. After that, analysis of the potential of economic complexity is carried out based on some differences reflected by information of external nation-export network and hidden network.