# Activist Success Predictor (Phase 67)
## Overview
ML model trained on historical activist campaign outcomes and governance scores to predict the success probability of activist campaigns and identify vulnerable targets.
## Requirements
```bash
pip install scikit-learn yfinance pandas numpy
```
**Note:** The module will run without scikit-learn but predictions will be disabled. Governance analysis and historical data will still work.
## CLI Commands
### 1. Predict Activist Success
Predict success probability for an activist campaign at a specific company.
```bash
python cli.py activist-predict AAPL
```
**Output:**
- Success probability (0-100%)
- Confidence level (High/Medium/Low)
- Governance metrics (score, ownership, board structure)
- Performance metrics (momentum, volatility, vs peers)
- Top influencing factors
- Risk factors and positive factors
### 2. Historical Campaign Analysis
Analyze historical activist campaign patterns and success rates.
```bash
python cli.py activist-history
```
**Output:**
- Total campaigns (2015-2024)
- Overall success rate
- Success rates by market cap tier
- Success rates by sector
- Common demands and their frequency
- Key success factors
### 3. Scan for Vulnerable Targets
Identify companies vulnerable to activist campaigns.
```bash
python cli.py activist-targets --sector Technology --min-cap 1000
```
**Options:**
- `--sector`: Filter by sector (Technology, Consumer, Industrial, Financial, Healthcare)
- `--min-cap`: Minimum market cap in millions (default: 1000)
**Output:**
- List of vulnerable targets ranked by success probability
- Governance scores and vulnerabilities
- Performance metrics
### 4. Governance Quality Score
Assess company governance quality and vulnerability.
```bash
python cli.py governance-score MSFT
```
**Output:**
- Governance score (0-100)
- Rating (Strong/Moderate/Weak)
- Ownership structure (insider/institutional)
- Board metrics (size, independence)
- Financial health metrics
- Performance vs market
- Vulnerabilities and strengths
## API Endpoints
### Predict Success
```
GET /api/v1/activist-predictor?action=predict&ticker=AAPL
```
### Historical Analysis
```
GET /api/v1/activist-predictor?action=history
```
### Scan Targets
```
GET /api/v1/activist-predictor?action=targets§or=Technology&min_cap=1000
```
### Governance Score
```
GET /api/v1/activist-predictor?action=governance-score&ticker=MSFT
```
## Data Sources
### Free APIs Used
1. **SEC EDGAR** - 13D/SC 13D activist filings
- Company CIK lookup
- Recent filings (13D, SC 13D, amendments)
- Filing dates and accession numbers
2. **Yahoo Finance** - Governance and market data
- Market cap, valuation (P/B ratio)
- Ownership structure (insider/institutional %)
- Financial metrics (ROE, debt/equity)
- Stock performance and volatility
- Historical price data
3. **S&P 500 Benchmark (SPY)** - Performance comparison
- 6-month returns
- Relative performance vs market
## ML Model Features
The Random Forest classifier uses 12 features:
1. **Valuation**: Market cap, Price-to-book ratio
2. **Performance**: ROE, 6-month momentum, volatility, vs peers
3. **Ownership**: Insider %, Institutional %
4. **Governance**: Composite governance score (0-100)
5. **Board Structure**: Size, independent director %
6. **Leverage**: Debt-to-equity ratio
### Training Data
Pre-trained on synthetic data based on historical patterns (2015-2024):
- 847 campaigns analyzed
- 58.8% overall success rate
- Success factors: Poor governance, undervaluation, low insider ownership
- Failure factors: Strong governance, fair value, high insider ownership
## Governance Score Calculation
Base score: 50
**Positive factors** (+):
- High institutional ownership (>70%): +15
- Strong ROE (>15%): +10
- Reasonable valuation (1.0 < P/B < 3.0): +5
**Negative factors** (-):
- High insider ownership (>25%): -15
- Poor ROE (<5%): -10
- Undervalued (P/B < 0.7): -10
Final score: Clamped to 0-100 range
## Success Factors
### High Success Indicators
- Governance score < 40
- Underperformance vs peers > 15%
- Insider ownership < 20%
- Price-to-book < 1.2
- Market cap < $10B
- Multiple institutional backers
### Low Success Indicators
- Governance score > 70
- Strong recent performance (>15%)
- High insider ownership (>25%)
- Large market cap (>$30B)
## Example Output
```json
{
"ticker": "AAPL",
"success_probability": 35.2,
"prediction": "Likely Resistance",
"confidence": "Medium",
"governance": {
"market_cap_m": 3999893.8,
"governance_score": 68.0,
"insider_ownership_pct": 1.8,
"institutional_ownership_pct": 65.2,
"price_to_book": 45.37,
"roe": 152.0
},
"performance": {
"momentum_6m_pct": 20.0,
"vs_peers_pct": 12.4,
"volatility_90d_pct": 22.6
},
"risk_factors": [
"Large market cap - harder to influence",
"Strong recent performance - less pressure for change"
],
"positive_factors": []
}
```
## Limitations
1. **Public Data Only**: No access to ISS/Glass Lewis proprietary governance ratings
2. **Board Estimation**: Board size and independence are estimated based on market cap
3. **Simplified Model**: Pre-trained on synthetic data; real-world training would require labeled campaign data
4. **No Real-time 13D Parsing**: SEC EDGAR filings are indexed, but full text parsing not implemented
5. **Limited Historical Data**: Yahoo Finance provides limited historical ownership data
## Future Enhancements
- Real-time 13D filing parsing for activist identity and intentions
- Integration with proxy fight data (Phase 69)
- Historical stock performance post-activism
- Activist track record analysis
- ISS/Glass Lewis API integration (paid)
- Deep learning model trained on actual campaign outcomes
- Natural language processing of proxy statements
## Technical Notes
- Requires ~10MB memory for ML model
- API timeout: 90 seconds (to allow for ML inference)
- Rate limiting: Respects Yahoo Finance rate limits
- Caching: No caching implemented; consider adding for production
## Testing
```bash
# Test governance score (no ML required)
python cli.py governance-score AAPL
# Test historical data (no ML required)
python cli.py activist-history
# Test ML prediction (requires scikit-learn)
python cli.py activist-predict AAPL
# Test target scanning (requires scikit-learn)
python cli.py activist-targets --sector Technology
```
