Automated Daily Trading System

An ML-powered stock prediction system that fetches real-time data from SimFin and generates next-day price movement predictions.

Overview

This project implements a full end-to-end pipeline:

1. ETL Pipeline: Downloads bulk stock data from SimFin, cleans it, engineers features (moving averages, volatility, lag returns), normalizes, and outputs ML-ready datasets using Polars
2. Model Training: Compares Logistic Regression and Random Forest classifiers, keeps the best model for each ticker
3. Web Application: Streamlit app with live predictions via the SimFin API and historical backtesting
4. Trading Strategy: Buy-and-Sell strategy based on model predictions, compared against a Buy-and-Hold baseline

Live Application

greatassignment.streamlit.app

Supported Tickers

MSFT, GOOG, AMZN, TSLA, NVDA

Project Structure

Home.py                  # Streamlit entry point (Home page)
pages/
  2_Go_Live.py           # Live prediction page
  3_Backtesting.py       # Trading strategy backtest page
src/
  pysimfin.py            # SimFin API wrapper (PySimFin class)
  etl.py                 # Standalone ETL script
  train_model.py         # Model training script (LR + RF comparison)
notebooks/
  Part_1.1_EDA_ETL.ipynb         # Data exploration and ETL pipeline
  Part_1.2_ML_Model.ipynb        # Model training and evaluation
  Part_1.3_Trading_Strategy.ipynb # Trading strategy backtest
models/                  # Trained model files (.joblib)
data/
  tickers_ml_ready/      # ML-ready Parquet files per ticker
  us-shareprices-daily.csv  # Raw SimFin bulk download (5 years of daily prices)
  us-companies.csv          # Company metadata from SimFin
docs/
  Executive_Summary.pdf  # Final executive summary with screenshots

Setup

1. Clone the repository
2. Create a conda environment: conda create -n my-assignment python=3.12
3. Install dependencies: pip install -r requirements.txt
4. Add your SimFin API key to a .env file: SIMFIN_API_KEY=your_key_here
5. Run the app: streamlit run Home.py

Tech Stack

• Polars for data processing (superior performance with large datasets and Parquet files)
• scikit-learn for ML models (Logistic Regression, Random Forest)
• SimFin for financial data (bulk downloads + REST API)
• Streamlit for the web application

Team

• Martin Sebastian Schneider Vaquero
• Francisco Javier Santiago Concha Bambach
• Aylin Yasgul
• Qiufeng Cai
• Bader Al Eisa