“Alternative” currencies such as Bitcoin and Gold, have had a phenomenal run in 2020 despite the pandemic.
We consider this period a unique opportunity to study and explore the investment applications of Long Short-Term Memory (LSTM) neural networks using Keras Python package to the directional classification problem of Bitcoin (XBT/USD) and Gold (Gold futures) daily price direction prediction.
We construct a neural network in Python using real data and we look at the LSTM network which involves an architectural modification to regular recurrent neural network.
According to recent groundbreaking findings on Bitcoin’s microstructure market, previous research up to now on Bitcoin price discovery omits the most important price-leading instruments and information flows.
Our work presents how these pioneering findings from new published research about derivatives products traded on three large unregulated crypto exchanges that strongly dominate price discovery within the bitcoin market could enhance our LSTM model accuracy. We use a similar framework with our approach for Gold.