Social Signal Sentiment-Based Prediction for Cryptocurrency Trading

Abstract

This project is keen on exploring the connection between the sentiment on a social-media platform about a cryptocurrency and the correlating price.

The approach has been to sample the sentiment from previously collected real-time tweets from Twitter. Afterwards, signals have been derived from these sentiment scores and a trading strategy was built. The system was designed to work in the background, store data in a Postgres database and trade on its own. It does this with the help of Heroku and a Scheduler, that checks every hour, if a trade should be made or not.

Insights of this project, like tweets, sentiment and trade metrics have been visualised with streamlit.

---

Live-Demo

The whole system is explained in the following video:

Further explanation about the visualisation (snapshot from 24th August, 2022):

(The videos are also found inside the GitHub Repo.)

---

Documentation

Table of Contents

• Abstract

• Introduction

• Research

  • Coin Comparison
  • Social Media Relevance
  • Focusing on Bitcoin

• Concept

  • Development Environment

• Data Acquisition

• Backend

• Sentiment

• Trading

• Visualisation

• Conclusion

• Appendices

First Chapter: Introduction >

(A PDF file containing all chapters is here. Unfortunately, the conversion from Markdown to PDF resulted in some faulty format. Better read here.)