Spaces:
Running
Running
File size: 16,897 Bytes
e6021a3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 | """
Advanced Mathematical Features for Pattern Recognition.
Institutional-grade feature engineering beyond standard technical indicators.
These features capture deep market microstructure that traditional indicators miss.
Features:
- Fourier Transform (spectral analysis, dominant frequencies)
- Fractal Dimension (market roughness via Higuchi method)
- Hurst Exponent (trending vs mean-reverting via R/S analysis)
- Shannon Entropy (market randomness/uncertainty)
- Autocorrelation Decay (momentum persistence)
- Volume-Price Correlation (smart money detection)
- Trend Strength Index (custom composite)
- Market Microstructure Features (tick patterns, gap analysis)
"""
from __future__ import annotations
import logging
from typing import Any, Dict, List
import numpy as np
import pandas as pd
logger = logging.getLogger(__name__)
class AdvancedFeatureEngine:
"""Compute advanced mathematical features from OHLCV data."""
def compute_all(self, df: pd.DataFrame) -> Dict[str, Any]:
"""
Compute all advanced features and return as a flat dict.
Returns both raw feature values and a features DataFrame
that can be used for ML model input.
"""
if df.empty or len(df) < 30:
return {"error": "Insufficient data (need 30+ bars)", "features": {}}
close = df["Close"].values.astype(float)
high = df["High"].values.astype(float)
low = df["Low"].values.astype(float)
volume = df["Volume"].values.astype(float) if "Volume" in df.columns else np.ones(len(close))
features: Dict[str, Any] = {}
# 1. Fourier Transform Analysis
try:
ft = self._fourier_analysis(close)
features.update(ft)
except Exception as e:
logger.debug("Fourier analysis failed: %s", e)
# 2. Fractal Dimension (Higuchi)
try:
features["fractal_dimension"] = self._higuchi_fractal_dimension(close)
except Exception as e:
logger.debug("Fractal dimension failed: %s", e)
# 3. Hurst Exponent
try:
features["hurst_exponent"] = self._hurst_exponent(close)
features["hurst_regime"] = (
"trending" if features["hurst_exponent"] > 0.55
else "mean_reverting" if features["hurst_exponent"] < 0.45
else "random_walk"
)
except Exception as e:
logger.debug("Hurst exponent failed: %s", e)
# 4. Shannon Entropy
try:
features["entropy_returns"] = self._shannon_entropy(close, bins=20)
features["entropy_volume"] = self._shannon_entropy(volume, bins=20)
except Exception as e:
logger.debug("Entropy failed: %s", e)
# 5. Autocorrelation Decay
try:
acf = self._autocorrelation_profile(close)
features.update(acf)
except Exception as e:
logger.debug("Autocorrelation failed: %s", e)
# 6. Volume-Price Correlation
try:
vpc = self._volume_price_analysis(close, volume)
features.update(vpc)
except Exception as e:
logger.debug("Volume-price analysis failed: %s", e)
# 7. Trend Strength Index
try:
features["trend_strength"] = self._trend_strength_index(close, high, low)
except Exception as e:
logger.debug("Trend strength failed: %s", e)
# 8. Gap Analysis
try:
gap = self._gap_analysis(df)
features.update(gap)
except Exception as e:
logger.debug("Gap analysis failed: %s", e)
# 9. Price Efficiency Ratio
try:
features["price_efficiency"] = self._price_efficiency_ratio(close)
except Exception as e:
logger.debug("Price efficiency failed: %s", e)
# 10. Kurtosis & Skewness of returns
try:
returns = np.diff(np.log(close + 1e-10))
if len(returns) >= 10:
features["return_skewness"] = float(pd.Series(returns).skew())
features["return_kurtosis"] = float(pd.Series(returns).kurtosis())
except Exception as e:
logger.debug("Moments failed: %s", e)
return features
def compute_feature_series(self, df: pd.DataFrame, window: int = 20) -> pd.DataFrame:
"""
Compute rolling advanced features as a DataFrame for ML training.
Returns one row per bar with multiple feature columns.
"""
result = df.copy()
close = df["Close"].values.astype(float)
volume = df["Volume"].values.astype(float) if "Volume" in df.columns else np.ones(len(close))
n = len(df)
hurst_vals = np.full(n, np.nan)
fractal_vals = np.full(n, np.nan)
entropy_vals = np.full(n, np.nan)
efficiency_vals = np.full(n, np.nan)
trend_vals = np.full(n, np.nan)
for i in range(window, n):
segment = close[i - window:i + 1]
vol_seg = volume[i - window:i + 1]
try:
hurst_vals[i] = self._hurst_exponent(segment)
except Exception:
pass
try:
fractal_vals[i] = self._higuchi_fractal_dimension(segment)
except Exception:
pass
try:
entropy_vals[i] = self._shannon_entropy(segment, bins=10)
except Exception:
pass
try:
efficiency_vals[i] = self._price_efficiency_ratio(segment)
except Exception:
pass
try:
high_seg = df["High"].values[i - window:i + 1].astype(float)
low_seg = df["Low"].values[i - window:i + 1].astype(float)
trend_vals[i] = self._trend_strength_index(segment, high_seg, low_seg)
except Exception:
pass
result["hurst_exponent"] = hurst_vals
result["fractal_dimension"] = fractal_vals
result["entropy"] = entropy_vals
result["price_efficiency"] = efficiency_vals
result["trend_strength"] = trend_vals
# Return-based distribution features (rolling)
returns = pd.Series(np.log(df["Close"] / df["Close"].shift(1)))
result["return_skew_20"] = returns.rolling(window).skew()
result["return_kurtosis_20"] = returns.rolling(window).apply(
lambda x: float(pd.Series(x).kurtosis()), raw=False
)
return result
# ββ Fourier Analysis βββββββββββββββββββββββββββββββββββββββββββββββββ
def _fourier_analysis(self, prices: np.ndarray, top_k: int = 5) -> Dict[str, Any]:
"""FFT spectral analysis of price series."""
log_prices = np.log(prices + 1e-10)
detrended = log_prices - np.linspace(log_prices[0], log_prices[-1], len(log_prices))
fft_vals = np.fft.rfft(detrended)
magnitudes = np.abs(fft_vals)
freqs = np.fft.rfftfreq(len(detrended))
# Skip DC component (index 0)
if len(magnitudes) > 1:
magnitudes = magnitudes[1:]
freqs = freqs[1:]
if len(magnitudes) == 0:
return {}
# Top dominant frequencies
top_indices = np.argsort(magnitudes)[-top_k:][::-1]
total_energy = np.sum(magnitudes ** 2)
dominant_periods = []
for idx in top_indices:
if freqs[idx] > 0:
period = 1.0 / freqs[idx]
energy_pct = (magnitudes[idx] ** 2) / total_energy * 100 if total_energy > 0 else 0
dominant_periods.append({
"period_bars": round(period, 1),
"energy_pct": round(energy_pct, 2),
})
# Spectral energy ratios
low_freq = magnitudes[:max(1, len(magnitudes)//4)]
high_freq = magnitudes[len(magnitudes)//4:]
low_energy = np.sum(low_freq ** 2)
high_energy = np.sum(high_freq ** 2)
return {
"fft_dominant_periods": dominant_periods[:3],
"fft_spectral_ratio": round(low_energy / (high_energy + 1e-10), 4),
"fft_total_energy": round(float(total_energy), 4),
}
# ββ Fractal Dimension (Higuchi) ββββββββββββββββββββββββββββββββββββββ
def _higuchi_fractal_dimension(self, x: np.ndarray, k_max: int = 10) -> float:
"""Higuchi fractal dimension β measures market roughness."""
n = len(x)
if n < k_max + 1:
k_max = max(2, n // 2)
lk = np.zeros(k_max)
for k in range(1, k_max + 1):
lm_sum = 0.0
for m in range(1, k + 1):
indices = np.arange(0, (n - m) // k) * k + m - 1
if len(indices) < 2:
continue
segment = x[indices.astype(int)]
length = np.sum(np.abs(np.diff(segment))) * (n - 1) / (k * len(segment))
lm_sum += length
lk[k - 1] = lm_sum / k if k > 0 else 0
# Fit log-log regression
valid = lk > 0
if np.sum(valid) < 2:
return 1.5 # default
ks = np.arange(1, k_max + 1)[valid]
log_k = np.log(1.0 / ks)
log_lk = np.log(lk[valid])
slope = np.polyfit(log_k, log_lk, 1)[0]
return round(float(slope), 4)
# ββ Hurst Exponent (R/S Analysis) ββββββββββββββββββββββββββββββββββββ
def _hurst_exponent(self, prices: np.ndarray) -> float:
"""
Rescaled range (R/S) Hurst exponent.
H > 0.5: trending, H < 0.5: mean-reverting, H β 0.5: random walk
"""
returns = np.diff(np.log(prices + 1e-10))
n = len(returns)
if n < 20:
return 0.5
max_k = min(n // 2, 100)
divisions = [d for d in range(10, max_k + 1, max(1, max_k // 20))]
if len(divisions) < 3:
return 0.5
rs_values = []
sizes = []
for d in divisions:
n_segments = n // d
if n_segments < 1:
continue
rs_list = []
for seg in range(n_segments):
segment = returns[seg * d:(seg + 1) * d]
mean_seg = np.mean(segment)
cumdev = np.cumsum(segment - mean_seg)
r = np.max(cumdev) - np.min(cumdev)
s = np.std(segment, ddof=1)
if s > 0:
rs_list.append(r / s)
if rs_list:
rs_values.append(np.mean(rs_list))
sizes.append(d)
if len(sizes) < 3:
return 0.5
log_sizes = np.log(np.array(sizes, dtype=float))
log_rs = np.log(np.array(rs_values, dtype=float))
slope = np.polyfit(log_sizes, log_rs, 1)[0]
return round(float(np.clip(slope, 0, 1)), 4)
# ββ Shannon Entropy ββββββββββββββββββββββββββββββββββββββββββββββββββ
def _shannon_entropy(self, data: np.ndarray, bins: int = 20) -> float:
"""Shannon entropy of data distribution. Higher = more random."""
if len(data) < 5:
return 0.0
hist, _ = np.histogram(data, bins=bins, density=True)
hist = hist[hist > 0]
if len(hist) == 0:
return 0.0
hist = hist / hist.sum() # normalize
return round(float(-np.sum(hist * np.log2(hist + 1e-12))), 4)
# ββ Autocorrelation Profile ββββββββββββββββββββββββββββββββββββββββββ
def _autocorrelation_profile(self, prices: np.ndarray) -> Dict[str, float]:
"""Compute autocorrelation at multiple lags."""
returns = np.diff(np.log(prices + 1e-10))
if len(returns) < 20:
return {}
result = {}
for lag in [1, 3, 5, 10, 20]:
if lag < len(returns):
acf = np.corrcoef(returns[lag:], returns[:-lag])[0, 1]
result[f"acf_lag_{lag}"] = round(float(acf), 4) if np.isfinite(acf) else 0.0
# Decay rate: how fast autocorrelation drops
acf_values = [result.get(f"acf_lag_{l}", 0) for l in [1, 5, 10, 20]]
result["acf_decay_rate"] = round(
float(np.polyfit(range(len(acf_values)), acf_values, 1)[0]), 6
)
return result
# ββ Volume-Price Analysis ββββββββββββββββββββββββββββββββββββββββββββ
def _volume_price_analysis(
self, prices: np.ndarray, volume: np.ndarray
) -> Dict[str, float]:
"""Analyze volume-price relationship for smart money detection."""
returns = np.diff(np.log(prices + 1e-10))
vol = volume[1:] # align with returns
if len(returns) < 10:
return {}
# Volume-return correlation
corr = np.corrcoef(returns, vol)[0, 1]
abs_corr = np.corrcoef(np.abs(returns), vol)[0, 1]
# Volume on up vs down days
up_mask = returns > 0
down_mask = returns < 0
avg_up_vol = np.mean(vol[up_mask]) if up_mask.sum() > 0 else 0
avg_down_vol = np.mean(vol[down_mask]) if down_mask.sum() > 0 else 0
vol_asymmetry = (avg_up_vol - avg_down_vol) / (avg_up_vol + avg_down_vol + 1e-10)
return {
"volume_return_corr": round(float(corr), 4) if np.isfinite(corr) else 0,
"volume_abs_return_corr": round(float(abs_corr), 4) if np.isfinite(abs_corr) else 0,
"volume_asymmetry": round(float(vol_asymmetry), 4),
}
# ββ Trend Strength Index βββββββββββββββββββββββββββββββββββββββββββββ
def _trend_strength_index(
self, close: np.ndarray, high: np.ndarray, low: np.ndarray
) -> float:
"""
Custom composite trend strength (0 = no trend, 1 = strong trend).
Combines ADX-like directional movement with price efficiency.
"""
n = len(close)
if n < 14:
return 0.5
# Directional movement
dm_plus = np.maximum(np.diff(high), 0)
dm_minus = np.maximum(-np.diff(low), 0)
# Nullify weaker direction
both = dm_plus > dm_minus
dm_plus[~both] = 0
dm_minus[both] = 0
# Smoothed (14-period average)
period = min(14, len(dm_plus))
avg_dm_plus = np.mean(dm_plus[-period:])
avg_dm_minus = np.mean(dm_minus[-period:])
total_dm = avg_dm_plus + avg_dm_minus
if total_dm == 0:
return 0.0
dx = abs(avg_dm_plus - avg_dm_minus) / total_dm
# Combine with efficiency
efficiency = self._price_efficiency_ratio(close)
return round(float((dx + efficiency) / 2), 4)
# ββ Price Efficiency Ratio βββββββββββββββββββββββββββββββββββββββββββ
def _price_efficiency_ratio(self, prices: np.ndarray) -> float:
"""
Kaufman Efficiency Ratio: net movement / total path length.
1.0 = perfect trend, 0.0 = pure noise.
"""
if len(prices) < 5:
return 0.5
net_change = abs(prices[-1] - prices[0])
total_path = np.sum(np.abs(np.diff(prices)))
return round(float(net_change / (total_path + 1e-10)), 4)
# ββ Gap Analysis βββββββββββββββββββββββββββββββββββββββββββββββββββββ
def _gap_analysis(self, df: pd.DataFrame) -> Dict[str, Any]:
"""Analyze price gaps for institutional activity detection."""
if len(df) < 10:
return {}
opens = df["Open"].values
prev_closes = df["Close"].shift(1).values
gaps = (opens[1:] - prev_closes[1:]) / (prev_closes[1:] + 1e-10)
gap_ups = gaps[gaps > 0.005]
gap_downs = gaps[gaps < -0.005]
return {
"gap_up_count_20": int(np.sum(gaps[-20:] > 0.005)) if len(gaps) >= 20 else 0,
"gap_down_count_20": int(np.sum(gaps[-20:] < -0.005)) if len(gaps) >= 20 else 0,
"avg_gap_size": round(float(np.mean(np.abs(gaps[-20:])) * 100), 4) if len(gaps) >= 20 else 0,
"max_gap_pct": round(float(np.max(np.abs(gaps[-20:])) * 100), 4) if len(gaps) >= 20 else 0,
}
# Module singleton
advanced_feature_engine = AdvancedFeatureEngine()
|