Learning Blockchain Delays: A Queueing Theory Approach

Despite the growing interest in cryptocurrencies, the delays incurred to confirm transactions are one of the factors that hamper the wide adoption of systems such as Bitcoin. Bitcoin transactions usually are confirmed in short periods (minutes), but still much larger than conventional credit card systems (seconds). In this work, we propose a framework encompassing machine learning and a queueing theory model to (i) identify which transactions will be confirmed; and (ii) characterize the confirmation time of confirmed transactions. The proposed queueing theory model accounts for factors such as the activity time of blocks and the mean time between transactions. We parameterize the model for transactions that are confirmed within minutes, suggesting that its integration into a more general framework is a step towards building scalability to Bitcoin.