FITE4801 | University of Hong Kong

By Chan Ka Ho, Ku Pak Ho, Lee Chun Hang

Supervised by Professor Chow Kam Pui

Bitcoin Pricing Insight with ML and NLP


Project Overview

Our investment goal is to achieve capital accumulation in the Bitcoin market through a multi-strategy approach. Leveraging data, technology, and research, we take smart risk with the information advantage and aim to deliver sustainable return regardless of market condition. We further break down our objectives as follows.

  • Objective 1: We use pricing model to capitalizes on short-term price fluctuation and long-term trend. Using machine learning, the model learns patterns from the past data and generate profits by buying low and selling high.
  • Objective 2: The price prediction model also incorporates views and opinions discussed among investors communities on social media and news outlets. We will first access the significance of investor sentiment. This alternative data can then be used to assist the model to predict price trend based on the results of sentiment analysis.
  • Objective 3: The model exploit price movements in response events, such as new Bitcoin project releases, regulatory announcements, technological updates, macroeconomic events. The pre-event strategy involves researching market rumours, assessing the likelihood of these rumours materializing, analyse sentiment. We act before its formal announcements and harvest profit if the rumour is true. At the time of the event announcement, price reaction model is used to adjust price prediction. For post-event strategy, the model continues to identify opportunities for trend continuation and mean reversion, as the market may initially overreact or underreact to the event.

For each model, the performance will be evaluated using common evaluation metrics used by researchers and the industry. Using the results, strengths and weakness of each method will be discussed, which will be used to determine the most effective model for our case. Lastly, we will create a dashboard to display the results of the chosen pricing model. This serves as a user interface for traders to get Bitcoin pricing insights.

Project Exhibition

{Project video}