StockSense-AI / backend /technical_signals.py
PROSTLE's picture
fix: clamp all sentiment/technical scores to strict (-0.99,0.99) β€” no boundary 1.0 or -1.0 values anywhere
786285f
Raw History Blame Contribute Delete
4.93 kB
"""
Technical signal scoring.
Converts raw indicator values into a directional score (-1 to +1).
Each signal has an individual sub-score; the aggregate is weighted.
"""
from __future__ import annotations
import numpy as np
import pandas as pd
def score_technical_signals(df: pd.DataFrame) -> dict:
"""
Analyze technical indicators and return a composite score.
Returns:
{
"score": float (-1 to +1),
"direction": "bullish" | "bearish" | "neutral",
"signals": {name: {"score": float, "value": float, "interpretation": str}},
}
"""
latest = df.iloc[-1]
signals = {}
# ── 1. RSI (weight: 25%) ──────────────────────────
rsi = float(latest.get("RSI", 50))
if rsi >= 75:
rsi_score = -0.8 # overbought β†’ bearish
rsi_interp = "Overbought"
elif rsi >= 65:
rsi_score = -0.3
rsi_interp = "Slightly overbought"
elif rsi <= 25:
rsi_score = 0.8 # oversold β†’ bullish
rsi_interp = "Oversold"
elif rsi <= 35:
rsi_score = 0.3
rsi_interp = "Slightly oversold"
else:
rsi_score = 0.0
rsi_interp = "Neutral"
signals["RSI"] = {"score": rsi_score, "value": round(rsi, 2), "interpretation": rsi_interp}
# ── 2. MACD Crossover (weight: 25%) ───────────────
macd = float(latest.get("MACD", 0))
macd_signal = float(latest.get("MACD_Signal", 0))
macd_diff = macd - macd_signal
if abs(macd_diff) < 0.01:
macd_score = 0.0
macd_interp = "Neutral"
elif macd_diff > 0:
macd_score = min(0.79, macd_diff * 10) # bullish crossover β€” capped at 0.79
macd_interp = "Bullish crossover"
else:
macd_score = max(-0.79, macd_diff * 10) # bearish crossover β€” capped at -0.79
macd_interp = "Bearish crossover"
signals["MACD"] = {"score": round(macd_score, 3), "value": round(macd_diff, 4), "interpretation": macd_interp}
# ── 3. SMA Trend (weight: 20%) ────────────────────
price = float(latest["Close"])
sma20 = float(latest.get("SMA_20", price))
sma50 = float(latest.get("SMA_50", price))
above_sma20 = price > sma20
above_sma50 = price > sma50
sma20_above_50 = sma20 > sma50 # golden cross
if above_sma20 and above_sma50 and sma20_above_50:
sma_score = 0.7 # strong uptrend
sma_interp = "Strong uptrend (Golden Cross)"
elif above_sma20 and above_sma50:
sma_score = 0.4
sma_interp = "Uptrend"
elif not above_sma20 and not above_sma50 and not sma20_above_50:
sma_score = -0.7 # strong downtrend
sma_interp = "Strong downtrend (Death Cross)"
elif not above_sma20 and not above_sma50:
sma_score = -0.4
sma_interp = "Downtrend"
else:
sma_score = 0.0
sma_interp = "Mixed trend"
signals["SMA_Trend"] = {"score": sma_score, "value": round(price - sma20, 2), "interpretation": sma_interp}
# ── 4. Bollinger Band Width (weight: 15%) ─────────
bb_width = float(latest.get("BB_Width", 0))
bb_upper = float(latest.get("BB_Upper", price))
bb_lower = float(latest.get("BB_Lower", price))
if price >= bb_upper * 0.99:
bb_score = -0.5 # near upper band β†’ reversal risk
bb_interp = "Near upper band"
elif price <= bb_lower * 1.01:
bb_score = 0.5 # near lower band β†’ bounce potential
bb_interp = "Near lower band"
else:
bb_score = 0.0
bb_interp = "Within bands"
signals["Bollinger"] = {"score": bb_score, "value": round(bb_width, 4), "interpretation": bb_interp}
# ── 5. ATR Volatility (weight: 15%) ───────────────
atr = float(latest.get("ATR", 0))
atr_pct = (atr / price * 100) if price > 0 else 0
if atr_pct > 3.0:
atr_score = -0.3 # high volatility = risk
atr_interp = "High volatility"
elif atr_pct > 2.0:
atr_score = -0.1
atr_interp = "Moderate volatility"
else:
atr_score = 0.1
atr_interp = "Low volatility"
signals["ATR"] = {"score": atr_score, "value": round(atr_pct, 2), "interpretation": atr_interp}
# ── Weighted composite ────────────────────────────
weights = {"RSI": 0.25, "MACD": 0.25, "SMA_Trend": 0.20, "Bollinger": 0.15, "ATR": 0.15}
composite = sum(signals[k]["score"] * weights[k] for k in weights)
composite = max(-0.99, min(0.99, composite))
if composite > 0.15:
direction = "bullish"
elif composite < -0.15:
direction = "bearish"
else:
direction = "neutral"
return {
"score": round(composite, 3),
"direction": direction,
"signals": signals,
}