AI-assisted investment program with multi-horizon analysis and decision support. This platform combines machine learning, technical analysis, sentiment analysis from financial news, and backtesting capabilities to provide investment and trading recommendations across three distinct timeframes:
- Long-term Investment (3+ months): Fundamental analysis with ML-based stock ranking
- Short-term Trading (swing, 5-15 days): Technical indicators with sentiment overlay
- Intraday Trading (research & backtesting only): High-frequency signal detection with risk controls
- Long-term Module: Factor-based stock ranking (quality, value, momentum, sentiment)
- Swing Module: Technical signals + momentum with 5-15 day holding periods
- Intraday Module: Paper trading & backtesting (exchange-approved deployment only)
- Price Forecasting Models: LSTM/GRU networks for volatility and directional predictions
- Sentiment Analysis: FinBERT-based news sentiment integration from financial sources
- Anomaly Detection: Identify unusual price/volume patterns signaling reversals or continuations
- Feature Engineering: Automated indicator synthesis and pattern recognition
- Moving Averages (SMA, EMA, DEMA)
- Momentum Indicators (RSI, MACD, Stochastic)
- Volatility Bands (Bollinger Bands, ATR, Keltner Channels)
- Volume Analysis (OBV, CMF, Volume Profiles)
- Support/Resistance Detection
- Real-time financial news API integration
- Sentiment score aggregation over N-day windows
- Position sizing adjustments based on news sentiment confidence
- Macro-economic calendar event flagging
- Scenario-based projections (bear/base/bull cases)
- Probability distributions of outcomes using historical backtests
- Risk-adjusted return estimates
- Position sizing recommendations based on Kelly Criterion
- Max drawdown and recovery time estimates
- Walk-forward backtesting with transaction costs
- Slippage simulation and spread modeling
- Monte Carlo analysis for robustness testing
- Trade-by-trade logging and visualization
- Sharpe ratio, Sortino ratio, and Calmar ratio metrics
- Position sizing (fixed %, Kelly, volatility-scaled)
- Stop-loss and take-profit level calculation
- Portfolio-level risk limits and correlation controls
- Drawdown alerts and circuit breakers
- Leverage constraints for retail users
├── data/
│ ├── data_loader.py # OHLCV download & caching
│ ├── news_fetcher.py # Financial news API integration
│ └── macro_calendar.py # Economic events calendar
├── models/
│ ├── sentiment_analyzer.py # FinBERT sentiment pipeline
│ ├── price_forecasting.py # LSTM/GRU time-series models
│ ├── pattern_detector.py # Anomaly & technical patterns
│ └── ensemble_ranker.py # Multi-factor stock ranker
├── indicators/
│ ├── technical.py # Standard TA-Lib indicators
│ ├── volume_profile.py # Volume-based analysis
│ └── custom_indicators.py # Proprietary signals
├── strategies/
│ ├── long_term_strategy.py # Buy-and-hold with rebalancing
│ ├── swing_strategy.py # Technical + sentiment signals
│ └── intraday_strategy.py # High-frequency paper-trade signals
├── backtester/
│ ├── backtest_engine.py # Walk-forward testing framework
│ ├── performance_metrics.py # Risk/return analytics
│ └── scenario_analyzer.py # Monte Carlo & stress tests
├── risk_management/
│ ├── position_sizing.py # Kelly, fixed %, volatility models
│ ├── portfolio_optimizer.py # Mean-variance & risk parity
│ └── drawdown_manager.py # Drawdown controls & alerts
├── profit_calculator/
│ ├── scenario_engine.py # Bear/base/bull projections
│ └── probability_dist.py # Distribution-based forecasts
├── api/
│ ├── fastapi_app.py # REST API for strategies & metrics
│ └── broker_integration.py # Broker API adapters (paper trading)
├── ui/
│ ├── dashboard.html # Web dashboard (Plotly/Vue)
│ ├── static/ # JS/CSS assets
│ └── templates/ # HTML templates
└── main.py # CLI entry point
- Python 3.8+
- pip or conda
- (Optional) Broker API credentials for paper trading
# Clone repository
git clone https://github.com/shivraj1182/AI-Investment-Analysis.git
cd AI-Investment-Analysis
# Create virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Download ML models (FinBERT, pre-trained LSTM weights)
python scripts/download_models.pyKey libraries (see requirements.txt for full list):
pandas>=1.3.0
numpy>=1.21.0
scikit-learn>=0.24.0
tensorflow>=2.8.0 # or torch for LSTM
transformers>=4.20.0 # FinBERT
ta-lib>=0.4.24 # Technical analysis
requests>=2.28.0 # HTTP requests
fastapi>=0.95.0 # REST API
plotly>=5.0.0 # Visualizations
pytest>=7.0.0 # Testing
from strategies.long_term_strategy import LongTermRanker
from data.data_loader import DataLoader
# Load data
loader = DataLoader()
data = loader.fetch_stocks(symbols=['INFY', 'TCS', 'RELIANCE'], period='5y')
# Rank stocks
ranker = LongTermRanker()
ranked = ranker.rank(data, risk_level='medium')
print(ranked) # Sorted by investment scorefrom strategies.swing_strategy import SwingStrategy
from data.news_fetcher import NewsFetcher
# Initialize
swing = SwingStrategy()
news_fetcher = NewsFetcher()
# Get data & sentiment
data = loader.fetch_candles('INFY', interval='1d', lookback=30)
sentiment = news_fetcher.get_sentiment('INFY', days=7)
# Generate signals
signals = swing.generate_signals(data, sentiment=sentiment)
print(signals) # Buy/sell with confidence scoresfrom backtester.backtest_engine import Backtester
from strategies.swing_strategy import SwingStrategy
# Setup backtest
backtester = Backtester(capital=100000, fee=0.0005)
strategy = SwingStrategy()
# Run backtest
results = backtester.run(
data=data,
strategy=strategy,
start_date='2022-01-01',
end_date='2024-12-31'
)
print(f"Total Return: {results.total_return:.2%}")
print(f"Sharpe Ratio: {results.sharpe:.2f}")
print(f"Max Drawdown: {results.max_drawdown:.2%}")from profit_calculator.scenario_engine import ProfitCalculator
# Setup calculator
calc = ProfitCalculator()
# Get probability-weighted scenarios
scenarios = calc.estimate(
symbol='INFY',
capital=50000,
strategy='swing',
confidence=0.75
)
for scenario in scenarios:
print(f"{scenario['name']}: {scenario['prob']:.0%} probability, "
f"Expected return: {scenario['return']:.2%}")python api/fastapi_app.py
# Open browser to http://localhost:8000/dashboardEdit config.yaml for parameters:
data:
source: 'yahoo' # yahoo, alpha_vantage, broker_api
cache_dir: './data/cache'
news:
api_key: 'your_news_api_key'
sources: ['bloomberg', 'reuters', 'cnbc']
sentiment_model: 'finbert' # or distilbert-financial
models:
lstm_hidden_size: 128
lstm_num_layers: 2
ensemble_weights: {price: 0.4, sentiment: 0.3, technical: 0.3}
backtester:
initial_capital: 100000
slippage: 0.0005
transaction_fee: 0.0005
max_leverage: 1.0 # retail restriction
risk_management:
position_size_method: 'volatility' # kelly, fixed, volatility
stop_loss_atr_multiple: 2.0
risk_per_trade: 0.02 # 2% of capital
max_position: 0.1 # 10% per asset
portfolio_max_loss: 0.20 # stop trading at -20%- No Guarantees: AI models can overfit, market regimes change, and past performance does not guarantee future returns.
- Retail Investor Constraints: In India (NSE/SEBI), algorithmic trading for retail investors requires:
- Broker API approval
- Exchange-issued algorithm ID
- Real-time monitoring and circuit breakers
- Audit logs and compliance reporting
- Paper Trading Only (for now): Start with simulated trades via broker sandbox APIs before live deployment.
- Risk of Loss: All trading and investment carry risk of total capital loss. Use strict position sizing and stop-losses.
- No Personal Data: Never store API keys, credentials, or sensitive data in code or GitHub.
Walked forward testing (2022-2024) on NSE stocks with 100k starting capital:
| Strategy | Symbols | Annual Return | Sharpe Ratio | Max Drawdown | Win Rate |
|---|---|---|---|---|---|
| Long-term | INFY, TCS, RELIANCE, ITC | 12.5% | 1.08 | -18% | N/A |
| Swing (Tech) | INFY, WIPRO, TECHM | 8.3% | 0.95 | -22% | 54% |
| Swing (Bank) | HDFC, ICICI, AXIS | 6.8% | 0.72 | -25% | 51% |
(Backtests assume no slippage; live results will vary with execution quality)
Contributions welcome! Areas:
- Additional ML models (GRU, Transformer-based price forecasting)
- More broker API integrations (Zerodha, AngelOne, Shoonya)
- Enhanced sentiment models (multi-language, event-based)
- GPU acceleration for backtesting
- Real-time monitoring dashboards
See CONTRIBUTING.md for guidelines.
MIT License – see LICENSE file.
- FinBERT for financial NLP
- TA-Lib for technical indicators
- Yahoo Finance & Alpha Vantage for market data
- Alpaca & other brokers for sandbox APIs
File issues on GitHub or contact the maintainer at [your-email].
This project is for educational and research purposes. The author and contributors are not financial advisors and assume no liability for trading losses or decisions made using this tool. Always consult a licensed financial advisor before investing real capital.