This repository contains the code used to execute the experiments described in the paper Predicting Cryptocurrencies Market Phases through On-Chain Data Long-Term Forecasting. Code is written in Python and using the Keras framework.
Extensive experiments have been executed testing stochastic processes (SARIMAX) and DL models (CNN and RNN) over six on-chain time series collected from Glassnode: New and Active Addresses, Block Height, Fees, Hash Rate, Spent Output Profit Ratio (SOPR).
Models have been trained on an Intel(R) Core(TM) i5-5257U CPU (2 cores per CPU).
To run the experiments as is, clone this repository and follow instructions:
1. Download the metrics from Glassnode
2. Put the data in the same folder of the Jupyter notebooks
3. Run all the cells of the Jupyter notebook
If you want to use additional datasets, you need to change only the first few cells where there is the import of the data.
Bruno Casella [email protected]
Lorenzo Paletto [email protected]