Econophysics from modern methods in mathematical science and theoretical physics (June 3rd 2019 - )
Analyzing big data on the basis of mathematical methods is now making a big impact not only on fundamental sciences but also on the human society. The purpose of the present econophysics (EcoP) WG is to focus on business to business (BtoB) transaction networks and to analyze their structure from the approaches developed in theoretical physics and mathematics.
In the following, we list several previous and on-going studies by the members.
1. Community detection
A complex network can be decomposed into sub-structures called communities. We apply a community detection method based on the theory of random walk, and are trying to find hidden communities characterized beyond simple industry-types.
2. Time evolution of complex network
We simulate the time evolution of complex BtoB networks by using an agent-based model. In particular, we are trying to extract the decision making policy of the companies from the real BtoB network data. Once we develop a realistic model, we will investigate the stabilization mechanism of the BtoB network, as well as the effects of loan and bankrupt of companies.
3. topology of complex network
To find a new indicator of the complex BtoB networks, we introduce the topological data analysis (TDA) based on the theory of persistent homology. TDA can detect n-dimensional “holes” in the dataset as well as their stability. We find that not only the direct application of TDA to the raw BtoB dataset but also to the reduced BtoB dataset after the community detection has a possibility to provide useful information .
Other than the above items, we have recently started (i) application of machine leaning for making growth forecast of companies and (ii) development of a method of dimensional reduction applicable to large BtoB networks.
- Yoshimasa Hidaka (RIKEN iTHEMS and Nishina Center)
- Takahiro Doi (RIKEN Nishina Center)