/stock-machine-learning-project

This project is for Chulalongkorn University's Pattern Recognition 2024 Class

Primary LanguageJupyter NotebookMIT LicenseMIT

Stock Return Prediction

A pattern recognition project for predicting stock return from historical data using multiple machine learning models and techniques.

See the slide Here.

Dataset: S&P500, Yahoo Finance - yfinance

Outline

  • Data Preprocessing
  • Feature Engineering
    • Technical Indicators
      • Relative Strength Index (RSI)
      • Bollinger Bands
      • Average True Range (ATR)
      • Moving Average Convergence/Divergence (MACD)
      • Momentum
      • Lagged Return
  • Data Splitting
    • Training and Testing Set: 2014 - 2023 (9y)
    • Trading Evaluation: 2023 - 2024 (1y)
  • Model
    • Baseline: Naive Forecast
    • Neural Network: Architecture 1 (Timeseries) & 2 (TS+Exogenous)
      • Long Short Term Memory (LSTM)
      • Gated Recurrent Unit (GRU)
    • Classical: Regression & Classification
      • Linear regression
      • Logistic regression
      • Support Vector
      • Random Forest
      • Extreme Gradient Boosting
      • K-Nearest Neighbor
  • Hyper parameters Tuning
    • Grid Search
    • Random Search
  • Trading
    • Baseline Strategy: Buy and Hold
    • Equal Weight portfolio
    • Sharpe Ratio
  • Analysis
    • Feature Importance

Contributors