MVLSC Home · Issue Contents · Forthcoming Papers
Analysis of Bitcoin Price Prediction Using Random Forests Method with On-Chain Data
Nava-Solís, Ángel Roberto and Rodríguez-García, Martha Del Pilar
This research contributes to the literature by empirically demonstrating the effectiveness of the random forest method and its variants, incorporating technical and on-chain metrics, as well as neural networks, for Bitcoin price prediction. Our research focuses on extracting meaningful information from large amounts of bitcoin prices, such as price and volume, from January 2015 to September 2, 2023. The Root Mean Square Error (RMSE) and the confusion matrix are used to evaluate the predictive level of the models. The model with the highest accuracy and consistency is the Random Forest model, with on-chain hash rate metrics of $769.61 USD and confusion matrix accuracy of 86% versus neural networks with $2761.69 USD and 17%, respectively. This research contributes to the literature by empirically demonstrating the use of the random forest method with on-chain data is better than the other methods analyzed.
Keywords: Cryptocurrencies, machine learning, finance market, on-chain metrics, artificial neural networks