crypto_trade

A service to project intraday spot price for crypto currency pairs (like BTC/USDT, TRON/USDT, TRON/BTC)

Summary

METIY concept for Crypto Trade architecture:

• Mine: this is a Docker-application that runs an Airflow to repeatedly collect a market data from OKX API server and write it to a postrges database
• Expose: a short Dash-application for visualization of a market data and estimations
• Train: this is an MLFlow image with a set of Catboost, Pytorch training pipelines
• Inference: a python-module for making predictions based on underlying ML-model
• Yield: an alert system or recommendation system for making trades [need more research]

Mining module

I used okx as the main datasource and platform for data-mining

Documentation: https://www.okx.com/docs-v5/en/

Data:

Data point	Type	Description	Requests limit	Reference to doc
Tickers	collection	the latest price snapshot, best bid/ask price, and trading volume in the last 24 hours	20 requests per 2 seconds	https://www.okx.com/docs-v5/en/#rest-api-market-data-get-tickers
Instruments	dict	List of instruments with open contracts	20 requests per 2 seconds	https://www.okx.com/docs-v5/en/#rest-api-public-data-get-instruments
System time	time	API server time	10 requests per 2 seconds	https://www.okx.com/docs-v5/en/#rest-api-public-data-get-system-time

Exposure module

TBD

Training module

TBD

Inference module

TBD

Yield module

TBD