Author Name | IKEDA Yuichi (Kyoto University) / AOYAMA Hideaki (Faculty Fellow, RIETI) / HATSUDA Tetsuo (RIKEN) / HIDAKA Yoshimasa (Kyoto University) / SHIRAI Tomoyuki (Kyushu University) / SOUMA Wataru (Rissho University) / IYETOMI Hiroshi (Rissho University) / Abhijit CHAKRABORTY (Indian Institutes of Science Education and Research / RIKEN) / FUJIHARA Akihiro (Chiba Institute of Technology) / NAKAYAMA Yasushi (SBI Financial and Economic Research Institute Co. Ltd) / ARAI Yuta (Reitaku University) / Krongtum SANKAEWTONG (Kyoto University) |
---|---|
Creation Date/NO. | December 2024 24-E-085 |
Research Project | Dynamics of Price in Crypto Assets and Real Economy and Their Underlying Complex Networks |
Download / Links |
Abstract
Realizing a cyber-physical economy requires dealing with the problems of the digital society that have arisen with the development of information technology. This study systematizes the mathematical basis for detecting anomalies for a dynamic graph, a network representation of relationships among nodes of crypto asset transactions and changes as time passes, based on graph theory, topology, and high-dimensional statistical analysis, to answer the three research questions: (1) Are there leading indicators of transactions that precede prices? (2) Is there a correlation between the velocity of circulation and prices? (3) Is there a herding phenomenon in the transaction network? Here, we define “anomaly” as large price fluctuations that affect transactions. The multiple methods above are applied to dynamic graphs during higher priced periods of crypto asset transactions to estimate individual anomaly indicators. We verify the effectiveness of the various anomaly detection methods by answering the three research questions for a major crypto asset. Finally, we propose a concept for an anomaly detection AI that estimates a comprehensive anomaly indicator by inputting various features from individual analysis methods.