This project aims to predict the directional movement of Bitcoin’s closing price as a binary classification task (up/down).
We will first address the trend and the cyclic behavior caused by the 4 year halving cycle. We will subtract the trend and seasonality from the BTC time series, making it approximately stationary.
We will model the resulting series based on a diverse set of features, including macroeconomic indicators (e.g., DXY index, interest rates, inflation, M2 money supply), stock indices (S&P 500, Nvidia), and prices of major cryptocurrencies. The goal is to explore whether external financial signals can enhance the predictability of short- to medium-term Bitcoin price movements using deep learning techniques.