File size: 13,266 Bytes
031a2d6 | 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 | from __future__ import annotations
from dataclasses import dataclass
from pathlib import Path
from typing import Dict, List
import numpy as np
import pandas as pd
@dataclass
class FeatureEngineerConfig:
input_path: str = "data/processed/merged_monthly.csv"
output_path_full: str = "data/features/features_monthly.csv"
output_path_long: str = "data/features/features_monthly_full_history.csv"
class FeatureEngineer:
"""
Builds model-ready monthly features from merged_monthly.csv.
Design:
- run() only creates engineered columns
- dataset splitting / dropna is done outside this class
- this preserves full history for alternative output datasets
Gold-focused additions:
- real_yield
- eurusd_level / gbpusd_level
- vix_squared
- gold_momentum_3m
- gold-specific lagged macro relationships
"""
def __init__(self, input_path: str) -> None:
self.input_path = Path(input_path)
self.df = self._load_data()
self.column_map = self._build_column_map()
def _load_data(self) -> pd.DataFrame:
if not self.input_path.exists():
raise FileNotFoundError(f"Input file not found: {self.input_path}")
df = pd.read_csv(self.input_path)
date_candidates = ["Date", "date", "DATE"]
date_col = next((c for c in date_candidates if c in df.columns), None)
if date_col is None:
raise ValueError("No date column found. Expected one of: Date, date, DATE")
df[date_col] = pd.to_datetime(df[date_col], errors="coerce")
if df[date_col].isna().all():
raise ValueError("Date column could not be parsed.")
df = df.dropna(subset=[date_col]).sort_values(date_col).reset_index(drop=True)
df = df.set_index(date_col)
for col in df.columns:
df[col] = pd.to_numeric(df[col], errors="coerce")
return df
@staticmethod
def _normalize(name: str) -> str:
return (
str(name)
.lower()
.replace("^", "")
.replace("/", "")
.replace("-", "")
.replace(" ", "")
.replace(".", "")
.replace("(", "")
.replace(")", "")
)
def _match_column(self, options: List[str]) -> str | None:
normalized = {self._normalize(c): c for c in self.df.columns}
for option in options:
norm_option = self._normalize(option)
if norm_option in normalized:
return normalized[norm_option]
return None
def _build_column_map(self) -> Dict[str, str]:
candidates = {
"spx": ["spx", "^spx_d", "sp500", "s&p500", "spx_close", "spx_price"],
"ndx": ["ndx", "^ndx_d", "nasdaq100", "nasdaq_100", "ndx_close", "ndx_price"],
"ftse": ["ftse100", "ftse", "ftse_100", "ftse_close", "ftse_price"],
"gold": ["gold", "xauusd", "gold_price", "xauusd_close"],
"btc": ["bitcoin", "btc", "btcusd", "btc_price", "btcusd_close"],
"eurusd": ["eurusd", "eur_usd", "eurusd_close", "eurusd_price"],
"gbpusd": ["gbpusd", "gbp_usd", "gbpusd_close", "gbpusd_price"],
"us2y": ["us2y_yield", "dgs2", "us2y", "us_2y", "treasury_2y"],
"us10y": ["us10y_yield", "dgs10", "us10y", "us_10y", "treasury_10y"],
"us_cpi": ["us_cpi", "cpiaucsl", "cpi_us", "us_cpi_index"],
"uk_cpi": ["uk_cpi", "cpi_uk_d", "cpi_uk", "uk_cpi_index"],
"hy_spread": ["us_hy_oas", "bamlh0a0hym2", "high_yield_spread", "hy_spread", "credit_spread"],
"vix": ["vix", "vixcls", "vix_level"],
"ecb": ["ecb_series"],
"dxy": ["dxy", "dx_y_nyb", "usd_index", "dxy_close"],
"qqq": ["qqq", "qqq_close", "qqq_ndx_proxy"],
"fed_funds": ["fed_funds", "fedfunds", "federal_funds_rate"],
"tips_10y": ["tips_10y", "dfii10", "tips_real_yield"],
"breakeven_10y": ["breakeven_10y", "t10yie", "breakeven_inflation"],
}
column_map: Dict[str, str] = {}
for logical_name, options in candidates.items():
found = self._match_column(options)
if found is not None:
column_map[logical_name] = found
missing_important = [k for k in ["spx", "ndx", "us2y", "us10y", "vix"] if k not in column_map]
if missing_important:
raise ValueError(
"Missing required columns for Phase 3: "
+ ", ".join(missing_important)
+ f"\nAvailable columns: {list(self.df.columns)}"
)
return column_map
def _log_return(self, col: str) -> pd.Series:
series = self.df[col]
return np.log(series / series.shift(1))
def _safe_add_return_feature(self, logical_name: str, feature_name: str) -> None:
if logical_name in self.column_map:
self.df[feature_name] = self._log_return(self.column_map[logical_name])
def compute_returns(self) -> None:
self._safe_add_return_feature("spx", "spx_return")
self._safe_add_return_feature("ndx", "ndx_return")
self._safe_add_return_feature("ftse", "ftse_return")
self._safe_add_return_feature("gold", "gold_return")
self._safe_add_return_feature("btc", "btc_return")
self._safe_add_return_feature("eurusd", "eurusd_return")
self._safe_add_return_feature("gbpusd", "gbpusd_return")
def compute_rolling_return_features(self) -> None:
if "gold_return" in self.df.columns:
self.df["gold_return_3m"] = self.df["gold_return"].rolling(3).sum()
self.df["gold_momentum_3m"] = self.df["gold_return"].rolling(3).sum()
self.df["gold_momentum_6m"] = self.df["gold_return"].rolling(6).sum()
if "spx_return" in self.df.columns:
self.df["spx_return_3m"] = self.df["spx_return"].rolling(3).sum()
if "ndx_return" in self.df.columns:
self.df["ndx_return_3m"] = self.df["ndx_return"].rolling(3).sum()
def compute_volatility(self) -> None:
if "spx_return" in self.df.columns:
self.df["spx_vol_3m"] = self.df["spx_return"].rolling(3).std()
self.df["spx_vol_6m"] = self.df["spx_return"].rolling(6).std()
if "ndx_return" in self.df.columns:
self.df["ndx_vol_3m"] = self.df["ndx_return"].rolling(3).std()
self.df["ndx_vol_6m"] = self.df["ndx_return"].rolling(6).std()
if "gold_return" in self.df.columns:
self.df["gold_vol_3m"] = self.df["gold_return"].rolling(3).std()
self.df["gold_vol_6m"] = self.df["gold_return"].rolling(6).std()
if "vix" in self.column_map:
self.df["vix_level"] = self.df[self.column_map["vix"]]
self.df["vix_squared"] = self.df["vix_level"] ** 2
def compute_rates(self) -> None:
self.df["us2y_yield"] = self.df[self.column_map["us2y"]]
self.df["us10y_yield"] = self.df[self.column_map["us10y"]]
self.df["yield_spread"] = self.df["us10y_yield"] - self.df["us2y_yield"]
def compute_inflation(self) -> None:
if "us_cpi" in self.column_map:
us_cpi_col = self.column_map["us_cpi"]
self.df["us_cpi_yoy"] = self.df[us_cpi_col] / self.df[us_cpi_col].shift(12) - 1
if "uk_cpi" in self.column_map:
uk_cpi_col = self.column_map["uk_cpi"]
self.df["uk_cpi_yoy"] = self.df[uk_cpi_col] / self.df[uk_cpi_col].shift(12) - 1
def compute_credit_risk(self) -> None:
if "hy_spread" in self.column_map:
self.df["high_yield_spread"] = self.df[self.column_map["hy_spread"]]
def compute_ecb_features(self) -> None:
if "ecb" in self.column_map:
ecb_col = self.column_map["ecb"]
self.df["ecb_level"] = self.df[ecb_col]
self.df["ecb_yoy"] = self.df[ecb_col] / self.df[ecb_col].shift(12) - 1
def compute_fx_level_features(self) -> None:
if "eurusd" in self.column_map:
self.df["eurusd_level"] = self.df[self.column_map["eurusd"]]
if "gbpusd" in self.column_map:
self.df["gbpusd_level"] = self.df[self.column_map["gbpusd"]]
def compute_gold_macro_features(self) -> None:
"""
Gold-specific macro features.
These are high-impact additions for Phase 6.1+ gold modelling.
"""
if {"us10y_yield", "us_cpi_yoy"}.issubset(self.df.columns):
self.df["real_yield"] = self.df["us10y_yield"] - self.df["us_cpi_yoy"]
if "real_yield" in self.df.columns:
self.df["real_yield_lag1"] = self.df["real_yield"].shift(1)
self.df["real_yield_change_1m"] = self.df["real_yield"].diff(1)
if "vix_level" in self.df.columns:
self.df["vix_lag1"] = self.df["vix_level"].shift(1)
self.df["vix_change_1m"] = self.df["vix_level"].diff(1)
if "yield_spread" in self.df.columns:
self.df["yield_spread_lag1"] = self.df["yield_spread"].shift(1)
if "high_yield_spread" in self.df.columns:
self.df["high_yield_spread_lag1"] = self.df["high_yield_spread"].shift(1)
if "gold_return" in self.df.columns:
self.df["gold_return_lag1"] = self.df["gold_return"].shift(1)
self.df["gold_return_lag2"] = self.df["gold_return"].shift(2)
if "eurusd_level" in self.df.columns:
self.df["eurusd_level_lag1"] = self.df["eurusd_level"].shift(1)
if "gbpusd_level" in self.df.columns:
self.df["gbpusd_level_lag1"] = self.df["gbpusd_level"].shift(1)
def compute_dxy_features(self) -> None:
if "dxy" in self.column_map:
dxy_col = self.column_map["dxy"]
self.df["dxy_level"] = self.df[dxy_col]
self.df["dxy_return"] = np.log(self.df[dxy_col] / self.df[dxy_col].shift(1))
self.df["dxy_return_3m"] = self.df["dxy_return"].rolling(3).sum()
def compute_qqq_features(self) -> None:
if "qqq" in self.column_map:
qqq_col = self.column_map["qqq"]
self.df["qqq_return"] = np.log(self.df[qqq_col] / self.df[qqq_col].shift(1))
def compute_fed_features(self) -> None:
if "fed_funds" in self.column_map:
col = self.column_map["fed_funds"]
self.df["fed_funds_level"] = self.df[col]
self.df["fed_funds_change_1m"] = self.df[col].diff(1)
self.df["fed_funds_change_3m"] = self.df[col].diff(3)
def compute_tips_breakeven_features(self) -> None:
if "tips_10y" in self.column_map:
col = self.column_map["tips_10y"]
self.df["tips_10y_level"] = self.df[col]
self.df["tips_10y_change_1m"] = self.df[col].diff(1)
if "breakeven_10y" in self.column_map:
col = self.column_map["breakeven_10y"]
self.df["breakeven_10y_level"] = self.df[col]
self.df["breakeven_10y_change_1m"] = self.df[col].diff(1)
# Forward-looking real yield: TIPS yield is the market-implied real rate
# (more precise than us10y_yield - us_cpi_yoy which is backward-looking)
if "tips_10y_level" in self.df.columns:
self.df["real_yield_tips"] = self.df["tips_10y_level"]
self.df["real_yield_tips_lag1"] = self.df["tips_10y_level"].shift(1)
self.df["real_yield_tips_change_1m"] = self.df["tips_10y_level"].diff(1)
def compute_stress_features(self) -> None:
if "vix_level" in self.df.columns:
vix_mean_12 = self.df["vix_level"].rolling(12).mean()
vix_std_12 = self.df["vix_level"].rolling(12).std()
self.df["vix_spike"] = (
self.df["vix_level"] > (vix_mean_12 + 2 * vix_std_12)
).astype(int)
if "spx" in self.column_map:
spx_price = self.df[self.column_map["spx"]]
spx_roll_max = spx_price.cummax()
self.df["spx_drawdown"] = (spx_price - spx_roll_max) / spx_roll_max
self.df["spx_max_dd_6m"] = self.df["spx_drawdown"].rolling(6).min()
if "ndx" in self.column_map:
ndx_price = self.df[self.column_map["ndx"]]
ndx_roll_max = ndx_price.cummax()
self.df["ndx_drawdown"] = (ndx_price - ndx_roll_max) / ndx_roll_max
self.df["ndx_max_dd_6m"] = self.df["ndx_drawdown"].rolling(6).min()
if "gold" in self.column_map:
gold_price = self.df[self.column_map["gold"]]
gold_roll_max = gold_price.cummax()
self.df["gold_drawdown"] = (gold_price - gold_roll_max) / gold_roll_max
self.df["gold_max_dd_6m"] = self.df["gold_drawdown"].rolling(6).min()
def run(self) -> None:
self.compute_returns()
self.compute_rolling_return_features()
self.compute_volatility()
self.compute_rates()
self.compute_inflation()
self.compute_credit_risk()
self.compute_ecb_features()
self.compute_fx_level_features()
self.compute_gold_macro_features()
self.compute_stress_features()
self.compute_dxy_features()
self.compute_qqq_features()
self.compute_fed_features()
self.compute_tips_breakeven_features() |