Understanding the driving mechanisms in Temporal Networks based on bitcoin transaction data

Complex Network theory applied to many real-world systems has proven to be a powerful tool to analyze and solve problems in complex systems. The Static Network approach is a developed field with lots of existing theories and models. However, real-world complex systems are not static, they evolve over time, change in size, create new connections etc. Approximating them as static is not always appropriate and sometimes it is possible to lose the crucial information by neglecting the time steps of interaction in the network. Temporal Network approach aims to incorporate temporal information into the complex networks and provides a fairer analysis of complex systems.
Currently, the field of Temporal Network is still emerging and lacks understanding of the underlying driving mechanisms. Using the empirical data of real-world complex systems is a good way to explore the structure of temporal networks, derive distinct features and construct models and theories based on those observations.
We chose the data set to investigate the temporal network structure according to the following criteria – it should be a real-world system with well-defined physical meaning of nodes, and precise time steps of nodes interaction. Bitcoin transaction data satisfies these conditions and moreover, it is readily available as open source data that was previously pre-processed into a user-friendly format. Time period for the analysis covers 4 year of transactions – from January 2009 till January 2013.
From the data we constructed the temporal, directed, evolving, weighted network. First, we performed static and temporal network analysis to explore the main properties and observe their dynamics over time. The main goal of our research is to investigate into the mechanistic properties of the time-evolving network. Exact time steps of the nodes’ (users’) interaction allowed us to investigate empirically the attachment and growth mechanisms of the network. We found that in bitcoin network the attachment mechanism has two main trends – preference to attach to new nodes with small degree and at the same time the preference to attach to the nodes with higher degree. Also, we have obtained the information of how network grows and how existing (old) nodes in system interact with each other over time. Based on the results it was possible to derive the model to simulate the synthetic evolving temporal network with tunable parameters that allow to reproduce the main properties of the system.