Title: Predicting the price of Bitcoin by its transaction network
Authors: Marcell Tamas Kurbucz - University of Pannonia (Hungary) [presenting]
Abstract: Studies on the Bitcoin transaction network have increased rapidly in recent years, but still, little is known about the networks influence on Bitcoin prices. The goals are twofold: to determine the predictive power of the transaction networks most frequent edges on the future price of Bitcoin and to provide an efficient technique for applying this untapped dataset in day trading. A complex method consisting of single-hidden layer feedforward neural networks (SLFNs) is used. The presented method achieved an accuracy of approximately 60.05\% during daily price movement classifications, despite only considering a small subset of edges.