A sustainable Bitcoin blockchain network through introducing dynamic block size adjustment using predictive analytics

Abstract

Bitcoin is the largest cryptocurrency in the market, which uses blockchain technology to bring in features like decentralization, anonymity, and trust. However, it still struggles with broader adaptation due to long verification times and high transaction fees. As a result, it is lagging behind competitors. We need to provide faster confirmations to tackle these issues while ensuring stable earnings for the miners. However, it is challenging to increase the block sizes or decrease the average block creation time without affecting the stability and security of the network. To address this conundrum, firstly, an optimization problem is formulated where the objective is to increase the transaction count in every cycle. Based on that, a comprehensive learning framework is developed to solve the formulated problem since the issue is intractable and hard to solve in polynomial time. The proposed learning framework includes (i) implementing a viable data-driven infrastructure with a machine learning (ML) root, (ii) training learning models with efficient generalization capability, and (iii) predicting the ideal block size in every block generation cycle. Our concept uses extreme gradient boost (XGB) as its core algorithm, which analyzes nine attributes associated with the Bitcoin network. These network-allied data points assist the model in creating an adaptive block size in the blockchain. XGB, trained using the last four years of real-world data, can predict block sizes with a 63.41% accuracy. The model ensures an all-around positive change in Bitcoin with a 12.29% increase in block size, a 13.45% increase in transaction fee (USD), and a 14.88% increase in transaction approval rate and transaction count, thus addressing the long wait time and broader adaption issue.