ecs-github-actions

Title

Thalassa trading tool: predicting the volatility of a cryptocurrency.

Problem

Experience traders' perception of risk (volatility) will affect their decision-making process, and hence their expected returns if they do not accurately infer future risk.

Solution

To help traders in their decision-making process, we build an application to predict in real-time a cryptocurrency's volatility and its regime.

Technical details

• By using data engineering, we gather, clean, and deliver high-frequency data (the order book) to feed models in a pipeline.
• By using statistics, we process the data to a frequency of seconds and construct financial features from the order book data. We employ the method of principal component analysis as a feature reduction.- By using machine learning, we train, validate and test both a time series model that uses as features the historical volatility and financial data to predict volatility in the next 30 seconds, and a gaussian mixture model to predict its regime.
• By using cloud platforms, we create and deploy a webpage to inform traders about the expected volatility and its regime.

Key technologies

Websocket, Pandas, Scikit-Learn, Streamlit, Doker, Google Cloud Platform.

Install requirements

pip install -r requirements.txt

To run the website locally

streamlit run Thalassa_Regime_Classifier/app.py