Prediction of Bitcoin Transactions Included in the Next Block

Abstract

This paper proposes a method to predict transactions that are likely to be included in the next block from the mempool of unconfirmed transactions in the Bitcoin network. To implement the proposed method, we applied machine learning to the transactions data collected from the Bitcoin network and divided our implementation into the following three objects: Data Collector; Data Preprocessor; and Analyzer. We used the random forest classifier algorithm because the problem of predicting the likelihood of a transaction to be included in the next block is a binary classification problem. We evaluated the performance of our model by comparing transactions in the mempool against transaction published in the next two blocks mined at the time of our experiments. For both blocks, our model has a prediction accuracy of more than 80% and a minimal false negative error. The analysis of transaction inclusion in the next block is fundamental as it could drive the price of Bitcoin or signify the properties of a given transaction such as illegal or legal.