Datasets:
Tasks:
Tabular Classification
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
economics
quantitative-finance
causal-inference
macroeconomics
housing-economics
market-microstructure
License:
File size: 15,021 Bytes
ebcde1f | 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 | """Leakage-safe empirical features and labels from normalized trades alone.
Trade availability time is the decision clock. Feature windows contain only the
current trade and earlier trades in the same ``(symbol, continuity_id)`` segment.
Future labels advance by trade ID inside that segment and preserve their explicit
information end so purged time-series evaluation can treat overlap correctly.
"""
from __future__ import annotations
import polars as pl
from microstructure.config import FeatureConfig
from microstructure.research.features import (
ResearchDataError,
TemporalAudit,
TemporalLeakageError,
)
_GROUP = ["symbol", "continuity_id"]
_REQUIRED_TRADES = frozenset(
{
"symbol",
"continuity_id",
"trade_id",
"available_ts_ns",
"event_ts_ns",
"price",
"quantity",
"aggressor_side",
}
)
def _require_columns(frame: pl.DataFrame, required: frozenset[str], table: str) -> None:
missing = sorted(required.difference(frame.columns))
if missing:
raise ResearchDataError(f"{table} is missing required columns: {missing}")
if frame.is_empty():
raise ResearchDataError(f"{table} must not be empty")
def _validate_config(config: FeatureConfig) -> None:
windows = (*config.trade_windows, config.intensity_window, config.volatility_window)
if not config.trade_windows or any(window < 1 for window in windows):
raise ResearchDataError("trade-only feature windows must be nonempty and positive")
if config.label_horizon_events < 1:
raise ResearchDataError("trade-only label horizon must be positive")
def _validate_normalized_trades(trades: pl.DataFrame) -> pl.DataFrame:
_require_columns(trades, _REQUIRED_TRADES, "normalized trades")
invalid = trades.filter(
pl.col("symbol").is_null()
| pl.col("continuity_id").is_null()
| pl.col("trade_id").is_null()
| pl.col("available_ts_ns").is_null()
| pl.col("event_ts_ns").is_null()
| (pl.col("available_ts_ns") < pl.col("event_ts_ns"))
| (~pl.col("price").cast(pl.Float64).is_finite())
| (pl.col("price") <= 0)
| (~pl.col("quantity").cast(pl.Float64).is_finite())
| (pl.col("quantity") <= 0)
| (~pl.col("aggressor_side").cast(pl.String).str.to_lowercase().is_in(["buy", "sell"]))
)
if not invalid.is_empty():
raise ResearchDataError(
"normalized trades require segment identity, observable timing, positive finite values, "
"and buy/sell aggressor side"
)
duplicates = trades.group_by(*_GROUP, "trade_id").len().filter(pl.col("len") > 1)
if not duplicates.is_empty():
raise ResearchDataError("trade IDs must be unique within each continuity segment")
ordered = trades.sort([*_GROUP, "trade_id"])
backwards = ordered.filter(
pl.col("available_ts_ns") < pl.col("available_ts_ns").shift(1).over(_GROUP)
)
if not backwards.is_empty():
raise ResearchDataError(
"available_ts_ns must be nondecreasing by trade_id within each continuity segment"
)
return ordered
def build_trade_only_features(trades: pl.DataFrame, config: FeatureConfig) -> pl.DataFrame:
"""Build causal rolling trade features on the availability-time clock."""
_validate_config(config)
ordered = _validate_normalized_trades(trades)
prepared = (
ordered.with_columns(
pl.col("event_ts_ns").alias("market_event_ts_ns"),
pl.col("available_ts_ns").alias("decision_ts_ns"),
pl.col("available_ts_ns").alias("feature_cutoff_ts_ns"),
pl.col("trade_id").alias("decision_trade_id"),
pl.col("trade_id").alias("decision_sequence"),
pl.col("continuity_id").alias("feature_continuity_id"),
pl.when(pl.col("aggressor_side").str.to_lowercase() == "buy")
.then(1.0)
.otherwise(-1.0)
.alias("trade_sign"),
pl.lit(1, dtype=pl.Int64).alias("_trade_observation"),
)
.with_columns(
(pl.col("quantity") * pl.col("trade_sign")).alias("signed_trade_quantity"),
pl.col("price").shift(1).over(_GROUP).alias("_previous_trade_price"),
pl.col("available_ts_ns").first().over(_GROUP).alias("_segment_start_ts_ns"),
pl.col("trade_id").cum_count().over(_GROUP).alias("history_trades"),
pl.concat_str(
["symbol", "continuity_id", pl.col("trade_id").cast(pl.String)],
separator=":",
).alias("sample_id"),
)
.with_columns(
pl.when(pl.col("_previous_trade_price").is_not_null())
.then((pl.col("price") / pl.col("_previous_trade_price")).log())
.otherwise(0.0)
.alias("log_trade_return_1")
)
)
expressions: list[pl.Expr] = []
windows = sorted(set((*config.trade_windows, config.intensity_window)))
for window in windows:
signed = (
pl.col("signed_trade_quantity")
.rolling_sum(window_size=window, min_samples=1)
.over(_GROUP)
)
volume = pl.col("quantity").rolling_sum(window_size=window, min_samples=1).over(_GROUP)
count = (
pl.col("_trade_observation").rolling_sum(window_size=window, min_samples=1).over(_GROUP)
)
window_start = pl.coalesce(
[
pl.col("decision_ts_ns").shift(window - 1).over(_GROUP),
pl.col("_segment_start_ts_ns"),
]
)
elapsed_seconds = (pl.col("decision_ts_ns") - window_start) / 1_000_000_000.0
expressions.extend(
[
signed.alias(f"signed_trade_volume_w{window}"),
volume.alias(f"trade_volume_w{window}"),
pl.when(volume > 0)
.then(signed / volume)
.otherwise(None)
.alias(f"trade_imbalance_w{window}"),
count.cast(pl.Float64).alias(f"trade_count_w{window}"),
pl.when(elapsed_seconds > 0)
.then(count / elapsed_seconds)
.otherwise(0.0)
.cast(pl.Float64)
.alias(f"trade_intensity_w{window}"),
]
)
volatility_window = config.volatility_window
expressions.append(
pl.col("log_trade_return_1")
.pow(2)
.rolling_sum(window_size=volatility_window, min_samples=1)
.over(_GROUP)
.sqrt()
.alias(f"realized_volatility_w{volatility_window}")
)
warmup = max((*config.trade_windows, config.intensity_window, config.volatility_window))
return (
prepared.with_columns(expressions)
.with_columns(
(pl.col("history_trades") >= warmup).alias("feature_ready"),
pl.col("decision_ts_ns").alias("max_feature_source_ts_ns"),
pl.col("decision_trade_id").alias("max_feature_source_trade_id"),
)
.drop("_trade_observation", "_previous_trade_price", "_segment_start_ts_ns")
.sort(["decision_ts_ns", "symbol", "continuity_id", "decision_trade_id"])
)
def add_future_trade_labels(frame: pl.DataFrame, horizon_trades: int) -> pl.DataFrame:
"""Attach strictly subsequent trade-price labels without crossing a gap."""
if horizon_trades < 1:
raise ResearchDataError("trade label horizon must be at least one trade")
_require_columns(
frame,
frozenset(
{
"symbol",
"continuity_id",
"decision_ts_ns",
"decision_trade_id",
"price",
}
),
"trade feature frame",
)
labeled = (
frame.sort([*_GROUP, "decision_trade_id"])
.with_columns(
pl.col("price").shift(-horizon_trades).over(_GROUP).alias("_target_trade_price"),
pl.col("decision_ts_ns")
.shift(-horizon_trades)
.over(_GROUP)
.alias("_target_trade_ts_ns"),
pl.col("decision_trade_id")
.shift(-horizon_trades)
.over(_GROUP)
.alias("_target_trade_id"),
pl.col("continuity_id")
.shift(-horizon_trades)
.over(_GROUP)
.alias("_target_continuity_id"),
)
.with_columns(
(
pl.col("_target_trade_price").is_null()
| (pl.col("_target_trade_id") <= pl.col("decision_trade_id"))
| (pl.col("_target_trade_ts_ns") < pl.col("decision_ts_ns"))
| (
(pl.col("_target_trade_ts_ns") == pl.col("decision_ts_ns"))
& (pl.col("_target_trade_id") <= pl.col("decision_trade_id"))
)
| (pl.col("_target_continuity_id") != pl.col("continuity_id"))
).alias("right_censored")
)
.with_columns(
pl.when(~pl.col("right_censored"))
.then((pl.col("_target_trade_price") / pl.col("price")).log())
.otherwise(None)
.alias("future_trade_return"),
pl.when(~pl.col("right_censored"))
.then(pl.col("_target_trade_price"))
.otherwise(None)
.alias("future_trade_price"),
pl.when(~pl.col("right_censored"))
.then(pl.col("_target_trade_ts_ns"))
.otherwise(None)
.alias("label_information_end_ts_ns"),
pl.when(~pl.col("right_censored"))
.then(pl.col("_target_trade_id"))
.otherwise(None)
.alias("label_information_end_trade_id"),
pl.when(~pl.col("right_censored"))
.then(pl.col("_target_continuity_id"))
.otherwise(None)
.alias("label_continuity_id"),
pl.lit(horizon_trades, dtype=pl.Int64).alias("label_horizon_trades"),
pl.col("decision_ts_ns").alias("label_start_ts_ns"),
pl.col("decision_trade_id").alias("label_start_trade_id"),
)
.with_columns(
pl.when(pl.col("future_trade_return").is_null())
.then(None)
.when(pl.col("future_trade_return") > 0)
.then(1)
.when(pl.col("future_trade_return") < 0)
.then(-1)
.otherwise(0)
.cast(pl.Int8)
.alias("future_trade_direction"),
pl.when(pl.col("future_trade_return").is_null())
.then(None)
.otherwise((pl.col("future_trade_return") > 0).cast(pl.Int8))
.alias("future_trade_up"),
)
.drop(
"_target_trade_price",
"_target_trade_ts_ns",
"_target_trade_id",
"_target_continuity_id",
)
.sort(["decision_ts_ns", "symbol", "continuity_id", "decision_trade_id"])
)
validate_trade_only_temporal_contract(labeled)
return labeled
def build_trade_only_research_frame(trades: pl.DataFrame, config: FeatureConfig) -> pl.DataFrame:
"""Build the complete causal trade-only feature and future-label frame."""
features = build_trade_only_features(trades, config)
return add_future_trade_labels(features, config.label_horizon_events)
def validate_trade_only_temporal_contract(frame: pl.DataFrame) -> TemporalAudit:
"""Fail closed when trade feature lineage or label timing is noncausal."""
required = frozenset(
{
"symbol",
"continuity_id",
"decision_ts_ns",
"decision_trade_id",
"feature_cutoff_ts_ns",
"feature_continuity_id",
"max_feature_source_ts_ns",
"max_feature_source_trade_id",
"right_censored",
"future_trade_return",
"future_trade_price",
"future_trade_direction",
"future_trade_up",
"label_start_ts_ns",
"label_start_trade_id",
"label_information_end_ts_ns",
"label_information_end_trade_id",
"label_continuity_id",
}
)
_require_columns(frame, required, "trade-only research frame")
future_feature = frame.filter(
(pl.col("feature_continuity_id") != pl.col("continuity_id"))
| (pl.col("max_feature_source_ts_ns") > pl.col("feature_cutoff_ts_ns"))
| (
(pl.col("max_feature_source_ts_ns") == pl.col("feature_cutoff_ts_ns"))
& (pl.col("max_feature_source_trade_id") > pl.col("decision_trade_id"))
)
)
if not future_feature.is_empty():
raise TemporalLeakageError("trade feature lineage extends beyond its decision cutoff")
uncensored = ~pl.col("right_censored")
invalid_label = frame.filter(
(pl.col("label_start_ts_ns") != pl.col("decision_ts_ns"))
| (pl.col("label_start_trade_id") != pl.col("decision_trade_id"))
| (
uncensored
& (
pl.col("future_trade_return").is_null()
| pl.col("future_trade_price").is_null()
| pl.col("future_trade_direction").is_null()
| pl.col("future_trade_up").is_null()
| pl.col("label_information_end_ts_ns").is_null()
| pl.col("label_information_end_trade_id").is_null()
| (pl.col("label_continuity_id") != pl.col("continuity_id"))
| (pl.col("label_information_end_trade_id") <= pl.col("decision_trade_id"))
| (pl.col("label_information_end_ts_ns") < pl.col("decision_ts_ns"))
| (
(pl.col("label_information_end_ts_ns") == pl.col("decision_ts_ns"))
& (pl.col("label_information_end_trade_id") <= pl.col("decision_trade_id"))
)
)
)
| (
pl.col("right_censored")
& (
pl.col("future_trade_return").is_not_null()
| pl.col("future_trade_price").is_not_null()
| pl.col("future_trade_direction").is_not_null()
| pl.col("future_trade_up").is_not_null()
| pl.col("label_information_end_ts_ns").is_not_null()
| pl.col("label_information_end_trade_id").is_not_null()
| pl.col("label_continuity_id").is_not_null()
)
)
)
if not invalid_label.is_empty():
raise TemporalLeakageError(
"trade labels are not strictly future, gap-local, and censor-safe"
)
censored_rows = frame.filter(pl.col("right_censored")).height
return TemporalAudit(
rows=frame.height,
labeled_rows=frame.height - censored_rows,
right_censored_rows=censored_rows,
continuity_segments=frame.select("symbol", "continuity_id").unique().height,
)
__all__ = [
"add_future_trade_labels",
"build_trade_only_features",
"build_trade_only_research_frame",
"validate_trade_only_temporal_contract",
]
|