An advanced stock price prediction system that leverages an ensemble of deep learning models to forecast future stock prices. The system combines three sophisticated neural network architectures:
- LSTM (Long Short-Term Memory): Specializes in capturing long-term dependencies in time series data
- GRU (Gated Recurrent Unit): Provides efficient sequence modeling with reduced complexity
- Hybrid CNN-LSTM: Combines convolutional layers for pattern recognition with LSTM layers for temporal analysis
- Real-time stock data fetching using yfinance
- Comprehensive technical indicators:
- Moving Averages (SMA, EMA)
- RSI (Relative Strength Index)
- MACD (Moving Average Convergence Divergence)
- Bollinger Bands
- Stochastic Oscillator
- ATR (Average True Range)
- ROC (Rate of Change)
- OBV (On-Balance Volume)
- Momentum
- Volatility metrics
- Dynamic model weight adjustment based on MAE and R² scores
- Interactive Streamlit UI with:
- Model performance metrics
- Training history visualization
- Future price predictions
- Technical analysis charts
- Advanced preprocessing with separate scaling for:
- Price-related features (open, high, low, close, volume)
- Technical indicators
- Support for 20+ major stock symbols
- Model persistence with automatic loading/saving
- Early stopping and learning rate reduction for better training
- Hyperparameter optimization using Optuna
- Two LSTM layers (64 and 32 units)
- L2 regularization (0.01)
- Batch normalization after each layer
- Dropout (0.2) after LSTM layers, (0.1) after dense layer
- Dense layers (16 units with ReLU, 1 unit for prediction)
- Adam optimizer (learning rate: 0.001)
- Two GRU layers (64 and 32 units)
- L2 regularization (0.01)
- Batch normalization after each layer
- Dropout (0.2) after GRU layers, (0.1) after dense layer
- Dense layers (16 units with ReLU, 1 unit for prediction)
- Adam optimizer (learning rate: 0.001)
- CNN layer (32 filters, kernel size 3)
- MaxPooling1D (pool size 2)
- Two LSTM layers (32 and 16 units)
- L2 regularization (0.01)
- Batch normalization after each layer
- Dropout (0.2) after LSTM layers, (0.1) after dense layer
- Dense layers (16 units with ReLU, 1 unit for prediction)
- Adam optimizer (learning rate: 0.001)
- Clone the repository:
git clone https://github.com/Aayush-05/Voila.git
cd Voila- Install required packages:
pip install -r requirements.txt- Run the Streamlit app:
streamlit run app.py-
Select a stock symbol from the dropdown menu
-
Choose training parameters:
- Training start date
- Prediction end date
- Force retrain model (optional)
- Auto-tune hyperparameters (optional)
- Training data years
- LSTM/GRU units (32-256)
- Dropout rate (0.1-0.5)
- Learning rate (0.0001-0.01)
- Batch size (16-128)
-
Click "Train Model and Predict" to start the process
stockprediction/
├── app.py # Main Streamlit application
├── model.py # Model architecture and training logic
├── data_preprocessing.py # Data fetchi
5331
ng and preprocessing
├── visualization.py # Plotting and visualization functions
├── config.py # Configuration parameters
├── requirements.txt # Project dependencies
└── saved_models/ # Directory for saved model files
- Python 3.8+
- TensorFlow 2.x
- Streamlit
- yfinance
- pandas
- numpy
- scikit-learn
- plotly
- ta (Technical Analysis library)
- optuna (for hyperparameter optimization)
- joblib (for model persistence)
- yfinance for providing stock data
- TensorFlow team for the deep learning framework
- Streamlit team for the web framework
- Technical Analysis library (ta) for indicators