Summary
In this chapter, we explored the analysis of price time series for BTC in a market that operates continuously, exhibits high volatility, and can experience exaggerated reactions to news events.
We began by familiarizing ourselves with the fundamental concepts of time series analysis and introduced traditional models such as ARIMA and Auto ARIMA. For our use case, we transformed our price dataset into the stationary form and learned to apply the models to it. Lastly, we incorporated an exogenous variable such as the news into our model. This external information proved to be valuable, contributing to a reduction in the error metric we were tracking.
Furthermore, we delved into the LSTM model approach, which required us to restructure the dataset differently. This involved numerous modifications and adaptations to accommodate the specific requirements of the LSTM model, which ultimately performed better.
By employing a comprehensive range of techniques and incorporating...
