A deep learning web application for predicting stock market prices using multiple model architectures, built with Python, Keras and Streamlit.
pip install -r requirements.txt
streamlit run app.py
| Model | Description |
|---|---|
| LSTM | Long Short-Term Memory for sequence prediction |
| Bi-LSTM + Attention | Bidirectional LSTM with custom attention mechanism |
| Transformer | Attention-based architecture for time series |
- 📊 Real-time stock data fetching via Yahoo Finance
- 📉 Moving Average analysis (MA50, MA100, MA200)
- 🔮 Future price prediction (30, 100, 200 days)
- 📋 Model accuracy comparison (RMSE, MAE, R² Score)
- 🏆 Automatic best model selection
- Python
- Keras / TensorFlow
- Streamlit
- Yahoo Finance (yfinance)
- Pandas & NumPy
- Scikit-learn
- Matplotlib
├── app.py # Streamlit web app
├── StockUpdate2.ipynb # LSTM & Bi-LSTM training
├── StockUpdate2_with_Transformer.ipynb # Transformer model training
├── LSTM_model.keras # Saved LSTM model
├── bi_lstm_model.keras # Saved Bi-LSTM model
├── bi_lstm_attention_model.keras # Saved Bi-LSTM + Attention model
└── README.md
- Enter any stock symbol (e.g. GOOG, AAPL, TSLA)
- App fetches historical data from 2012–2022
- Models predict future prices
- Dashboard shows original vs predicted prices
- Accuracy comparison table shows best performing model
Jalal Abedin — LinkedIn | GitHub