Bitcoin Transaction Confirmation Time Prediction: A Classification View

With Bitcoin being universally recognised as the most popular cryptocurrency, more Bitcoin transactions are expected to be populated to the Bitcoin blockchain system. As a result, many transactions can encounter different confirmation delays. One of the most demanding requirements for users is to estimate the confirmation time of a newly submitted transaction. In this paper, we argue that it is more practical to predict the confirmation time as falling into a time interval rather than falling onto a specific timestamp. After dividing the future into a set of time intervals (i.e. classes), the prediction of a transaction’s confirmation can be considered as a classification problem. Consequently, a number of mainstream classification methods, including neural networks and ensemble learning models, are evaluated. For comparison, we also design a baseline classifier that considers only the transaction feerate. Experiments on real-world blockchain data demonstrate that ensemble learning models can obtain higher accuracy, while neural network models perform better on the f1-score, especially when more classes are used.