In a recent study, the N-HiTS (Neural Hierarchical Interpolation for Time Series) model, a deep learning framework, was employed to predict Bitcoin prices for the next 30 days using Onchain data from the past 180 days. The model's unique ability to decompose input data into hierarchical levels allows it to capture different temporal patterns effectively.

The N-HiTS model generates intermediate forecasts through an interpolation mechanism, which are then recursively refined for accuracy. This approach enables the model to capture both short-term fluctuations and long-term trends effectively.

The modeling and training were conducted using PyTorch, PyTorch Lightning, and PyTorch Forecasting libraries. The training data included 376 features taken from the cryptoquant platform. The predicted and actual prices after the training process for the validation data, as well as the forecast for the next 30 days, were presented.

This application of the N-HiTS model in the blockchain industry demonstrates the potential of deep learning frameworks in enhancing the accuracy of cryptocurrency price predictions.