●62 reads●MIT License
AI-DOT: Artificial Intelligence Driven Options Trading Bot
- ai in finance
- algorithmic trading
- ann model
- artificial intelligence
- automated trading system
- backtesting strategies
- broker api integration
- buy sell signals
- data
- deep learning
- finance-ml pipeline
- financial forecasting
- keras model
- machine learning
- model evaluation metrics
- neural networks
- options data analytics
- options trading
- predictive modeling
- python trading bot
- quant-finance
- quantitative trading
- realtime trading
- rnn lstm
- stock price prediction
- technical indicators
- tensorflow
- time series prediction
- yfinance
Table of contents
🤖 AI-DOT: Artificial Intelligence Driven Options Trading Bot
AI-DOT is an intelligent options trading bot developed using Artificial Neural Networks (ANN) and Recurrent Neural Networks (RNN). It automatically analyzes options data, generates call/put buy/sell signals, and integrates with broker APIs for real-time execution. This project demonstrates the full pipeline from data acquisition to model training, evaluation, signal generation, and live deployment.
📦 Dataset & Features
- Data Source: yFinance
- Scope:
- Historical stock prices
- Option chain data (call/put)
- Implied volatility, open interest
- Preprocessing:
- Handling NaNs and outliers (Box & Line plots)
- Short/long moving averages
- StandardScaler normalization
- Feature Types:
- Technical Indicators (MA, volatility, price patterns)
- Derived options data (IV, call/put ratios)
⚙️ Modeling Approach
- Algorithms Used:
- ANN: For regression-based price prediction
- RNN (LSTM): To capture temporal patterns
- Architecture:
- Dense layers with ReLU activation
- Regularization: L1/L2
- Optimizer: Adam (lr=0.0001)
- EarlyStopping & ReduceLROnPlateau used
- Training Details:
- Epochs: 600 (ANN), 200 (RNN)
- Loss: MAE, RMSE
📊 Model Evaluation
| Model | MAE | RMSE |
|---|---|---|
| ANN | 0.37 | 0.47 |
| RNN | 1.18 | 1.58 |
ANN outperformed RNN in both accuracy and consistency. RNN showed stable learning but struggled with prediction precision.
📈 Learning Curve
Training and validation loss decreased smoothly with no overfitting observed.
💹 Trading Signal Generation
The trained models were used to generate trading signals:
- Green: Buy Call Signal
- Red: Buy Put Signal
Signals were overlaid on live price charts for intuitive visualization and backtesting.
🧪 Testing Workflow
| Test Stage | Status |
|---|---|
| Data preprocessing and cleaning | ✅ Passed |
| ANN & RNN Model training | ✅ Passed |
| Model validation with MAE & RMSE | ✅ Passed |
| Prediction histogram and alignment | ✅ Passed |
| Signal generation from model outputs | ✅ Passed |
| Visual verification of trading points | ✅ Passed |
🧠 Key Insights
- ANN provided more precise and stable predictions than RNN.
- Both models generalized well on unseen data with minimal overfitting.
- Buy/sell signal alignment with price trends confirmed model robustness.
- Visual inspection helped refine strategies and filter noise.
🚀 Future Scope
- Integrate live news sentiment analysis to enhance prediction accuracy.
- Expand models using hybrid CNN-RNN or Transformer-based architectures.
- Deploy a Streamlit or Flask-based dashboard with real-time brokerage execution.
- Enable multi-asset support (index options, forex derivatives, etc.)
📝 Citation
Malaviya, A., Dhumale, N., Kotian, N., & Vala, J. (2024). AI-DOT: Artificial Intelligence Driven Options Trading Bot. Shah and Anchor Kutchhi Engineering College.
📂 Resources
- 💻 GitHub Repository
- 📈 Report PDF
- 🔗 Broker API Integration (planned in real-time module)