Stock Price Predictor

This project uses LSTM (Long Short-Term Memory) neural networks to predict stock prices based on historical data and forecast future trends.

Features

• Fetches historical stock data using yfinance
• Preprocesses data using MinMaxScaler
• Implements LSTM neural network for time series prediction
• Provides performance metrics (RMSE and R² score)
• Visualizes predictions vs actual prices
• Future price predictions with confidence intervals
• Docker support for easy deployment
• Live code editing through volume mounts

Web Demo

Example prediction showing historical data, current predictions, and future forecast with confidence intervals:

Local Installation

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

Docker Usage

Using Docker Compose (Recommended)

1. Build and start the container:

docker-compose up -d --build

2. View logs:

docker-compose logs -f

3. Stop the container:

docker-compose down

Using Docker Directly

1. Build the image:

docker build -t stock-predictor .

2. Run the container:

docker run -d \
  -v $(pwd)/data:/app/data \
  -v $(pwd)/models:/app/models \
  -v $(pwd)/app.py:/app/app.py \
  -p 8501:8501 \
  --name stock-predictor \
  stock-predictor

Model Configuration

The stock prediction model:

• Uses 60-day sequences for prediction
• LSTM architecture with two layers (50 units each)
• Dropout layers (0.2) for regularization
• Training/Testing split: 80%/20%
• Customizable training epochs (default: 50)
• Future prediction window: 30 days
• Confidence intervals based on prediction variance

Data Storage

• Trained models are saved in the models/ directory
• Historical data is cached in the data/ directory
• Both directories are mounted as Docker volumes for persistence
• app.py is mounted for live development

Default Settings

• Default stock: TSMC (2330.TW)
• Default start date: 2025-01-01
• End date: Current date
• Timezone: Asia/Taipei

Interactive Features

• Stock symbol input (Taiwan/US markets)
• Adjustable date range
• Configurable sequence length and epochs
• Real-time visualization updates
• Future trend predictions with uncertainty ranges

Output

The application displays:

• Training and Testing RMSE (Root Mean Square Error)
• R² scores for both training and testing sets
• Interactive plot with:
  • Historical prices (blue line)
  • Model predictions (red dashed line)
  • Future predictions (with confidence intervals)
• Training loss history visualization