Alpha-Agent

Overview

Features

• Multi-Agent Architecture: Specialized agents for data, prediction, trading, risk management, and sentiment analysis
• Alpha Factor Selection: Implementation of 101 alphas and custom factors with dynamic evaluation
• Machine Learning Models: LSTM, XGBoost, and Prophet models for price prediction
• Options Strategies: Black-Scholes pricing and advanced options strategy generation
• Polygon.io Integration: Real-time and historical data from Polygon.io API
• Backtesting Framework: Comprehensive backtesting with detailed performance metrics
• Agent Communication: Robust communication protocol inspired by metaGPT

Installation

# Clone the repository
git clone https://github.com/yourusername/alpha-agent.git
cd alpha-agent
Install dependencies
pip install -r requirements.txt
Set environment variables for API keys
export POLYG
export OPENAIAPIKEY="youropenaiapi_key"  # If using sentiment analysis

Configuration

Edit the configuration file at configs/config.yaml to customize:

• Default tickers to monitor
• Data update intervals
• Model hyperparameters
• Trading parameters
• Risk management settings

Usage

Basic Usage

# Start the Alpha-Agent system
from alphaagent.main import startsystem
Start the system with default configuration
start_system()
Or specify a custom configuration file
start_system(c)

Running Individual Agents

# Run the prediction agent in standalone mode
from agents.prediction_agent import PredictionAgent
from utils.communication.unified_communication import UnifiedCommunicationManager
Initialize communication manager
communicator = UnifiedCommunicationManager()
communicator.start()
Create and start prediction agent
predictionagent = PredictionAgent("predictionagent", communicator)
prediction_agent.start()

Running Backtests

# Run a backtest
from backtest.backtest_engine import BacktestEngine
from backtest.historicaldatafetcher import HistoricalDataFetcher
Fetch historical data
datafetcher = HistoricalDataFetcher(apikey="yourpolygonapi_key")
dataset = datafetcher.fetchcomplete_dataset(
    tickers=["AAPL", "MSFT", "GOOGL"],
    start_date="2022-01-01",
    end_date="2022-12-31"
)
Run backtest
backtester = BacktestEngine(initial_capital=100000)
results = backtester.runbacktest(dataset["stocks"], yourstrategy_function)
Plot results
backtester.plotportfolioperformance()
backtester.plottradeanalysis()

Evaluating Alpha Factors

# Evaluate alpha factors
from models.signals.alpha_factors import AlphaFactors
from utils.factorevaluation.factoranalyzer import FactorAnalyzer
Calculate alpha factors
alpha_factors = AlphaFactors()
dfwithalphas = alphafactors.calculatealphafactors(historicaldata)
Evaluate factor performance
analyzer = FactorAnalyzer()
metrics = analyzer.calculatefactormetrics(
    data=dfwithalphas,
    factors=["alpha1", "alpha12", "alpha101"],
    forwardreturnsperiods=[1, 5, 10]
)
Get best factors
bestfactors = analyzer.getbest_factors(metric="ic", n=5)
print(f"Best factors: {best_factors}")

Adding Custom Alpha Factors

from models.signals.alpha_factors import AlphaFactors
class ExtendedAlphaFactors(AlphaFactors):     def custom_momentum(self, df):         """Custom momentum alpha factor."""         return df['close'].pct_change(20)  # 20-day momentum
Use your custom implementation
my_factors = ExtendedAlphaFactors()
dfwithalphas = myfactors.calculatealpha_factors(data)

System Architecture

+------------------+    +------------------+    +-------------------+
|                  |    |                  |    |                   |
|    Data Agent    |<-->| Prediction Agent |<-->|   Trading Agent   |
|                  |    |                  |    |                   |
+------------------+    +------------------+    +-------------------+
        ^                       ^                       ^
        |                       |                       |
        v                       v                       v
+------------------+    +------------------+    +-------------------+
|                  |    |                  |    |                   |
|  Sentiment Agent |<-->| Communication    |<-->|    Risk Agent     |
|                  |    |    Manager       |    |                   |
+------------------+    +------------------+    +-------------------+

Evaluation Metrics

The system provides comprehensive evaluation metrics for both individual agents and the overall system:

• Prediction Accuracy: RMSE, MAE, Directional Accuracy
• Trading Performance: Returns, Sharpe, Sortino, Maximum Drawdown
• Factor Performance: IC, Turnover, Half-life, Factor correlation
• System Performance: Agent communication efficiency, runtime metrics

Documentation

Detailed documentation is available in the docs/ directory:

• System architecture and design: docs/README.md
• API reference documentation: docs/api_reference.md
• Factor analysis guide: docs/factor_analysis.md
• Backtesting guide: docs/backtesting.md

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.