Random Forest regression over technical indicators to predict a 5-day price range and generate market insights for the Single Ticker Audit view.
Usage
from stock_analyzer import StockAnalyzer, generate_market_analysis
analyzer = StockAnalyzer()
data = analyzer.calculate_technical_indicators(ohlcv_df)
model_info = analyzer.train_prediction_model(data, horizon=5)
prediction = analyzer.predict_price_range(model_info, current_price=150.0)
insights = generate_market_analysis(data, ticker="AAPL")
Requirements
Needs pandas and scikit-learn. Run with the repo-root venv, NOT bare python3:
.venv/bin/python -c "from stock_analyzer import StockAnalyzer, generate_market_analysis"
Verified: ./.venv/bin/python has both. Bare python3 does not, and the resulting
ModuleNotFoundError: No module named 'pandas' is the "skill seems broken" symptom,
not a code fault.
Indicators
| Category | Indicators |
|---|---|
| Moving averages | SMA(20), SMA(50), EMA(12), EMA(26) |
| Momentum | RSI(14), MACD + signal + histogram |
| Volatility | Bollinger Bands(20,2), ATR(14) |
| Volume | Volume SMA(20), Volume Ratio |
| Oscillators | Stochastic %K, %D |
ML features: lag (close/volume/returns 1–5d back), rolling mean/std (5/10/20d), price-vs-SMA20/50, 10/20d return volatility. Minimums: 50 bars to train, 30 after split.
Prediction output
{ 'expected': 152.30, 'low': 148.50, 'high': 156.10,
'expected_change_pct': 1.53, 'confidence': 0.72, 'horizon': 5 }
low/high = ±1 std dev across trees; confidence = inverse of relative
uncertainty (0–1).
Insight categories
Price movement · RSI overbought/oversold · MA alignment · Bollinger extremes ·
MACD direction & crossovers · volume conviction. Returned as an emoji-annotated
list, e.g. 💡 RSI at 28.5 suggests oversold — potential buying opportunity.
Integration
main.py → render_audit_view(): fetch ~6mo daily (yfinance) → indicators →
train (if enough data) → prediction card (Low/Expected/High) → insights panel.
Source
Full guide: STOCK_ANALYZER_README.md.