Five Minute Midas

The most famous King Midas is popularly remembered in Greek mythology for his ability to turn everything he touched into gold. - Wikipedia

• Read the Medium article on this project!

Overview

• Predicting profitable day trading positions by fitting a random forest classifier on historical minute-level stock price data
• Independent end-to-end machine learning project, from data collection to model deployment
• Try the Web App Demo!

Features

• Data collection from Yahoo Finance (SQLite3)
• Data transformation, feature engineering (pandas)
• ML model training, tuning and tracking (scikit-learn, MLflow)
• ML model deployment: API and web app (Flask, Streamlit)

Data Pipeline

Installation

• Use requirements.txt for the demo.
• Use requirements_full.txt for all scripts.

Methodology

• Minute-level price data is extracted, and filtered to those with Bullish RSI Divergence
• These filtered points and their respective profit/loss outcomes are used to train an ML classifier
• With the trained model, we can try to predict future profit/loss outcomes

Credits

• Price data extracted with the help of the yfinance library, created and maintained by Ran Aroussi and other contributors

Contact

Michael Ow @ LinkedIn