Analyses of complex networks of industries and formulations of the growing models

公司之间建立了各种各样的关系。这些关系不是独立的，而是依赖的。此外，这种关系也影响到公司。因此，为了了解企业之间的关系的行为，我调查了以企业为节点，以各种关系为纽带的网络。在这里，我使用了专利的合著和由企业员工撰写的论文的合著。人们发明了许多方法来分析网络。我选择了P模型。p模型是一种概率模型，可以检测概率重要的结构。在模型的基础上，笔者发现日本的四个重要产业分别具有不同的结构。例如，有一种结构，一个人专门发表论文或专利。作为另一种分析，我使用了交易和专利合著。这些关系比其他变量（如公司的部门或规模）的联系更紧密。这是由贝叶斯网络获得的，而不能通过普通最小二乘分析获得。

Companies create various kinds of relationships with each other. The relationships must be not independent but dependent. Also, the relationships affect with companies. Therefore, to know the behavior of the relationships of companies, I investigated networks where companies are nodes and various kinds of the relationships are links.Here, I used co-authorships of patents and co-authorships of papers that were written by company workers. Many methods have been invented to analyze networks. I chose p model from them. p model is a probabilistic model that can detect probabilistically significant structures. On the basis of the model, I found four important Japanese industries have different structures respectively. For example, there is a structure where a person publishes papers or patents specifically.As another analysis, I used transactions and patent co-authorships. These relationships are more strongly connected than other variables, such as sectors or scales of companies. This was acquired by Baysian networks and cannot be acquired by ordinally least square analyses.