Understanding Cryptocurrency Price Formation from Time Series of Local Blockchain Graph Features
Cuneyt Akcora, Department of Computer Science, University of Texas at Dallas, USA
Over the last couple of years, digital cryptocurrencies and the Blockchain technology that forms their basis have witnessed a flood of attention. With the emergence and rapid adoption of Blockchain and the associated cryptocurrencies understanding the network dynamics behind Blockchain technologies has emerged as an important research direction. Unlike other financial networks such as stock and currency trading, blockchains have the entire time series of interaction graph accessible to the public. This facilitates a thorough analysis of the network data with a time series approach. A natural question to ask is whether the network dynamics of a cryptocurrency impact its price in dollars. We show that on the one hand, time series of standard global graph features such as degree distribution are not enough to capture the network dynamics that impact the underlying cryptocurrency price. In contrast, multiple time series of persistent topological homologies can explain higher level interactions among nodes in Blockchain graphs and can be used to build more accurate price prediction models.