Trade-Agent

Trade-Agent is a project that applies Deep Q-Learning (DQN), a reinforcement learning technique, to predict and analyze stock prices using various sources of stock market data.

Features

• Deep Q-Learning Model: Utilizes DQN for stock price prediction and trading strategy optimization.
• Multiple Data Sources: Supports data acquisition from Yahoo Finance, Alpha Vantage, Quandl, and more.
• Technical Indicators: Integrates TA-Lib, pandas-ta, and other libraries for feature engineering.
• Visualization: Provides data visualization using Matplotlib, Seaborn, and Plotly.
• Modular Design: Easily extendable for new data sources and RL algorithms.

Model Architure

==========================================================================================
Layer (type:depth-idx)                   Output Shape              Param #
==========================================================================================
Sequential                               [1, 3]                    --
├─Linear: 1-1                            [1, 1024]                 61,440
├─LayerNorm: 1-2                         [1, 1024]                 2,048
├─ReLU: 1-3                              [1, 1024]                 --
├─Dropout: 1-4                           [1, 1024]                 --
├─Linear: 1-5                            [1, 512]                  524,800
├─LayerNorm: 1-6                         [1, 512]                  1,024
├─ReLU: 1-7                              [1, 512]                  --
├─Dropout: 1-8                           [1, 512]                  --
├─Linear: 1-9                            [1, 256]                  131,328
├─LayerNorm: 1-10                        [1, 256]                  512
├─ReLU: 1-11                             [1, 256]                  --
├─Dropout: 1-12                          [1, 256]                  --
├─Linear: 1-13                           [1, 128]                  32,896
├─LayerNorm: 1-14                        [1, 128]                  256
├─ReLU: 1-15                             [1, 128]                  --
├─Dropout: 1-16                          [1, 128]                  --
├─Linear: 1-17                           [1, 64]                   8,256
├─ReLU: 1-18                             [1, 64]                   --
├─Linear: 1-19                           [1, 3]                    195
==========================================================================================
Total params: 762,755
Trainable params: 762,755
Non-trainable params: 0
Total mult-adds (M): 0.76
==========================================================================================
Input size (MB): 0.00
Forward/backward pass size (MB): 0.03
Params size (MB): 3.05
Estimated Total Size (MB): 3.08
==========================================================================================

Installation

1. Clone the repository:

   git clone https://github.com/OpenVanguard/Trade-Agent.git
   cd Trade-Agent

2. Set up the Python environment (recommended: use stock_rl_env):

   python -m venv stock_rl_env
   source stock_rl_env/Scripts/activate  # On Windows
   # Or
   source stock_rl_env/bin/activate      # On Linux/Mac

3. Install dependencies:

   pip install -r requirements.txt

Usage

• Refer src/README.md. OR
• Run the main script:

  python src/main.py

• Explore data analysis and experiments in notebooks/dataAnalysis.ipynb.

Project Structure

• src/: Main source code for RL environment and training.
• data/: Storage for raw and processed stock data.
• notebooks/: Jupyter notebooks for exploratory analysis.
• stock_rl_env/: Python virtual environment.

Requirements

• Python 3.10+
• PyTorch
• Stable Baselines3
• Gym
• yfinance, pandas_datareader, alpha_vantage, quandl, investpy
• TA-Lib, pandas-ta, scikit-learn, matplotlib, seaborn, plotly

License

This project is licensed under the MIT License. See LICENSE for details.

Author

Virat Srivastava