Sparse Group Network effects for Bitcoin Blockchain

Department of Statistics & Applied Probability
National University of Singapore

Analysis of the blockchain transaction is necessary to understand the interactions of the users. It gives insight into the state of the network on a global level, providing implications for the inherent risk of an investment into Bitcoin (BTC). The analysis faces though a dimensionality problem since the dynamic dependence structure is complex yet of extremely sparse nature. We propose a Sparse Group Network AutoRegressive (SGNAR) model. We present a regu-larized estimator which copes both group and individual sparsity to investigate the essential dependence in the blockchain transactions. This allows us to detect active groups with influential impact on the global network. Underlying BTC network dynamic effects in year to year show signs for the blockchain being in an adoption phase. Effects are identified coming from Europe and North America, yet only in the recent years, while surprisingly Asia does not affect the transaction network.