Spaces:
Sleeping
Sleeping
File size: 18,994 Bytes
bef09da | 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 | """
Feature engineering for the AA Crew Sequences Bad Weather XGBoost model.
Pipeline:
1. Load raw BTS parquet files
2. Build airport-level weather risk features (by airport x month)
3. Construct sequence pairs: inbound (A→DFW) + outbound (DFW→B) same day/pilot window
4. Build pair-level features
5. Label pairs as high-risk (target)
6. Save feature matrix for model training
"""
import os
import numpy as np
import pandas as pd
RAW_DIR = os.path.join(os.path.dirname(__file__), "..", "data", "raw")
PROCESSED_DIR = os.path.join(os.path.dirname(__file__), "..", "data", "processed")
os.makedirs(PROCESSED_DIR, exist_ok=True)
# Thresholds
WEATHER_DELAY_THRESHOLD_MIN = 15 # minutes to count as a weather delay event
TURNAROUND_MAX_HRS = 4 # max hours between inbound arrival and outbound departure
TURNAROUND_MIN_HRS = 0.5 # min turnaround (shorter = invalid / not a real sequence)
# ---------------------------------------------------------------------------
# Step 1: Load all raw data
# ---------------------------------------------------------------------------
def load_raw() -> pd.DataFrame:
# Prefer all-carrier files over AA-only; fall back to AA-only if not present
all_files = sorted([
os.path.join(RAW_DIR, f)
for f in os.listdir(RAW_DIR)
if f.startswith("bts_all_dfw_") and f.endswith(".parquet")
])
aa_files = sorted([
os.path.join(RAW_DIR, f)
for f in os.listdir(RAW_DIR)
if f.startswith("bts_aa_dfw_") and f.endswith(".parquet")
])
# For each year use all-carrier if available, else AA-only
year_file = {}
for f in aa_files:
yr = f.split("_")[-1].replace(".parquet", "")
year_file[yr] = f
for f in all_files:
yr = f.split("_")[-1].replace(".parquet", "")
year_file[yr] = f # overrides AA-only
files = list(year_file.values())
if not files:
raise FileNotFoundError(f"No BTS parquet files found in {RAW_DIR}. Run download_bts.py first.")
print(f"Using files: {[os.path.basename(f) for f in sorted(files)]}")
# Load year-by-year and aggregate airport features to avoid OOM
frames = []
for f in sorted(files):
frames.append(pd.read_parquet(f))
df = pd.concat(frames, ignore_index=True)
del frames
df["FlightDate"] = pd.to_datetime(df["FlightDate"])
df["Month"] = df["FlightDate"].dt.month
df["Year"] = df["FlightDate"].dt.year
df["Season"] = df["Month"].map({
12: "winter", 1: "winter", 2: "winter",
3: "spring", 4: "spring", 5: "spring",
6: "summer", 7: "summer", 8: "summer",
9: "fall", 10: "fall", 11: "fall",
})
print(f"Loaded {len(df):,} flights across {df['Year'].nunique()} years")
return df
# ---------------------------------------------------------------------------
# Step 2: Airport-level weather risk features
# ---------------------------------------------------------------------------
def build_airport_features(df: pd.DataFrame) -> pd.DataFrame:
"""
Compute per-airport, per-month aggregates used as features.
We use ALL flights touching that airport (origin or dest) to get robust stats.
"""
# For each flight, the "airport of interest" for weather is the non-DFW end
inbound = df[df["Dest"] == "DFW"].copy()
inbound["airport"] = inbound["Origin"]
inbound["weather_delayed"] = (inbound["WeatherDelay"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN).astype(int)
inbound["weather_cancel"] = ((inbound["Cancelled"] == 1) & (inbound["CancellationCode"] == "B")).astype(int)
inbound["weather_delay_min"] = inbound["WeatherDelay"].fillna(0)
inbound["nas_delayed"] = (inbound["NASDelay"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN).astype(int)
outbound = df[df["Origin"] == "DFW"].copy()
outbound["airport"] = outbound["Dest"]
outbound["weather_delayed"] = (outbound["WeatherDelay"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN).astype(int)
outbound["weather_cancel"] = ((outbound["Cancelled"] == 1) & (outbound["CancellationCode"] == "B")).astype(int)
outbound["weather_delay_min"] = outbound["WeatherDelay"].fillna(0)
outbound["nas_delayed"] = (outbound["NASDelay"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN).astype(int)
combined = pd.concat([inbound, outbound], ignore_index=True)
agg = combined.groupby(["airport", "Month"]).agg(
total_flights=("weather_delayed", "count"),
weather_delay_rate=("weather_delayed", "mean"),
weather_cancel_rate=("weather_cancel", "mean"),
avg_weather_delay_min=("weather_delay_min", "mean"),
p75_weather_delay_min=("weather_delay_min", lambda x: x.quantile(0.75)),
p95_weather_delay_min=("weather_delay_min", lambda x: x.quantile(0.95)),
nas_delay_rate=("nas_delayed", "mean"),
).reset_index()
# Overall (across all months) per airport
overall = combined.groupby("airport").agg(
overall_weather_delay_rate=("weather_delayed", "mean"),
overall_weather_cancel_rate=("weather_cancel", "mean"),
overall_avg_weather_delay_min=("weather_delay_min", "mean"),
).reset_index()
agg = agg.merge(overall, on="airport", how="left")
print(f"Airport features: {len(agg):,} rows ({agg['airport'].nunique()} airports × months)")
return agg
# ---------------------------------------------------------------------------
# Step 3: Construct sequence pairs
# ---------------------------------------------------------------------------
def _parse_hhmm(series: pd.Series) -> pd.Series:
"""Convert HHMM integer (e.g. 1435) to minutes since midnight."""
s = series.fillna(0).astype(int).astype(str).str.zfill(4)
hours = s.str[:2].astype(int)
minutes = s.str[2:].astype(int)
return hours * 60 + minutes
def build_sequences(df: pd.DataFrame) -> pd.DataFrame:
"""
Build (airport_A, airport_B, date) aggregates without materializing
flight-level pairs. For each date, collapse inbound flights to unique
origins and outbound to unique destinations, then cross-join the airport
aggregates (~100 unique origins × ~100 unique destinations = ~10k rows/day)
instead of flight-level pairs (~400 × ~400 = 160k rows/day).
"""
inbound = df[(df["Dest"] == "DFW") & (df["Cancelled"] != 1)].copy()
outbound = df[(df["Origin"] == "DFW") & (df["Cancelled"] != 1)].copy()
inbound["arr_min"] = _parse_hhmm(inbound["ArrTime"])
outbound["dep_min"] = _parse_hhmm(outbound["DepTime"])
# Aggregate inbound flights to (date, airport_A): worst-case weather outcomes
ib_agg = (
inbound.groupby(["FlightDate", "Origin", "Month", "Season", "Year"])
.agg(
arr_min_earliest = ("arr_min", "min"),
arr_min_latest = ("arr_min", "max"),
weather_delay_A = ("WeatherDelay", lambda x: x.fillna(0).max()),
arr_delay_A = ("ArrDelay", lambda x: x.fillna(0).max()),
nas_delay_A = ("NASDelay", lambda x: x.fillna(0).max()),
late_aircraft_A = ("LateAircraftDelay",lambda x: x.fillna(0).max()),
)
.reset_index()
.rename(columns={"Origin": "airport_A"})
)
# Aggregate outbound flights to (date, airport_B): worst-case weather outcomes
ob_agg = (
outbound.groupby(["FlightDate", "Dest"])
.agg(
dep_min_earliest = ("dep_min", "min"),
dep_min_latest = ("dep_min", "max"),
weather_delay_B = ("WeatherDelay", lambda x: x.fillna(0).max()),
dep_delay_B = ("DepDelay", lambda x: x.fillna(0).max()),
late_aircraft_B = ("LateAircraftDelay",lambda x: x.fillna(0).max()),
)
.reset_index()
.rename(columns={"Dest": "airport_B"})
)
# Merge on date → cross-join of unique airports per day (~10k rows/day vs 160k)
pairs = ib_agg.merge(ob_agg, on="FlightDate", how="inner")
pairs = pairs[pairs["airport_A"] != pairs["airport_B"]].copy()
# Feasibility filter: at least one valid turnaround window exists
# (earliest outbound departs after earliest inbound arrives + min turnaround,
# and latest outbound departs before latest inbound arrives + max turnaround)
turnaround_mid = pairs["dep_min_earliest"] - pairs["arr_min_latest"]
feasible = (
(pairs["dep_min_latest"] >= pairs["arr_min_earliest"] + TURNAROUND_MIN_HRS * 60) &
(pairs["dep_min_earliest"] <= pairs["arr_min_latest"] + TURNAROUND_MAX_HRS * 60)
)
pairs = pairs[feasible].copy()
pairs["turnaround_min"] = turnaround_mid[feasible].clip(lower=0)
print(f"Constructed {len(pairs):,} airport-pair×date rows from {pairs['FlightDate'].nunique():,} dates")
return pairs
# ---------------------------------------------------------------------------
# Step 4: Build pair-level feature matrix
# ---------------------------------------------------------------------------
def build_feature_matrix(pairs: pd.DataFrame, airport_features: pd.DataFrame) -> pd.DataFrame:
"""
Join airport-level features onto each pair and compute pair-level derived features.
"""
feat = airport_features.rename(columns=lambda c: f"A_{c}" if c not in ("airport", "Month") else c)
feat_b = airport_features.rename(columns=lambda c: f"B_{c}" if c not in ("airport", "Month") else c)
# Join airport A features
pairs = pairs.merge(
feat.rename(columns={"airport": "airport_A"}),
on=["airport_A", "Month"], how="left"
)
# Join airport B features
pairs = pairs.merge(
feat_b.rename(columns={"airport": "airport_B"}),
on=["airport_B", "Month"], how="left"
)
# Pair-level derived features
pairs["pair_combined_weather_rate"] = (
pairs["A_weather_delay_rate"] * pairs["B_weather_delay_rate"]
)
pairs["pair_max_weather_rate"] = pairs[["A_weather_delay_rate", "B_weather_delay_rate"]].max(axis=1)
pairs["pair_min_weather_rate"] = pairs[["A_weather_delay_rate", "B_weather_delay_rate"]].min(axis=1)
pairs["pair_weather_rate_sum"] = pairs["A_weather_delay_rate"] + pairs["B_weather_delay_rate"]
pairs["pair_avg_weather_delay_min"] = (
pairs["A_avg_weather_delay_min"] + pairs["B_avg_weather_delay_min"]
) / 2
pairs["both_high_risk"] = (
(pairs["A_weather_delay_rate"] > pairs["A_weather_delay_rate"].quantile(0.75)) &
(pairs["B_weather_delay_rate"] > pairs["B_weather_delay_rate"].quantile(0.75))
).astype(int)
# Season dummies
pairs = pd.get_dummies(pairs, columns=["Season"], prefix="season", drop_first=False)
pairs["is_spring_summer"] = pairs.get("season_spring", 0) | pairs.get("season_summer", 0)
return pairs
# ---------------------------------------------------------------------------
# Step 5: Label sequences as high-risk
# ---------------------------------------------------------------------------
def label_sequences(pairs: pd.DataFrame) -> pd.DataFrame:
"""
A sequence is labeled 1 (bad) if:
- Either leg had a significant weather delay (>= threshold), OR
- Leg 1 arrival delay cascaded into leg 2 departure (late aircraft propagation)
"""
weather_A = pairs["weather_delay_A"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN
weather_B = pairs["weather_delay_B"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN
cascade = (
(pairs["arr_delay_A"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN) &
(pairs["late_aircraft_B"].fillna(0) >= WEATHER_DELAY_THRESHOLD_MIN)
)
pairs["target"] = (weather_A | weather_B | cascade).astype(int)
print(f"Label distribution: {pairs['target'].mean():.1%} high-risk sequences")
return pairs
# ---------------------------------------------------------------------------
# Step 6: Final feature selection and save
# ---------------------------------------------------------------------------
FEATURE_COLS = [
# Airport A features
"A_weather_delay_rate", "A_weather_cancel_rate", "A_avg_weather_delay_min",
"A_p75_weather_delay_min", "A_p95_weather_delay_min", "A_nas_delay_rate",
"A_overall_weather_delay_rate", "A_overall_avg_weather_delay_min",
# Airport B features
"B_weather_delay_rate", "B_weather_cancel_rate", "B_avg_weather_delay_min",
"B_p75_weather_delay_min", "B_p95_weather_delay_min", "B_nas_delay_rate",
"B_overall_weather_delay_rate", "B_overall_avg_weather_delay_min",
# Pair features
"pair_combined_weather_rate", "pair_max_weather_rate", "pair_min_weather_rate",
"pair_weather_rate_sum", "pair_avg_weather_delay_min", "both_high_risk",
# Temporal
"Month", "is_spring_summer", "turnaround_min",
# Season dummies (added dynamically)
]
def save_features(pairs: pd.DataFrame, year: int = None):
"""
Aggregate flight-level pairs to (airport_A, airport_B, Month, Year).
This is the right unit of analysis: we want to score pair × month combos,
not individual sequences. Reduces 21M rows → ~tens of thousands.
Features are stable airport-level stats; target is observed bad rate binarized.
"""
season_cols = [c for c in pairs.columns if c.startswith("season_")]
# Stable feature cols: take first value per group (they're the same within group)
static_feat_cols = [c for c in FEATURE_COLS + season_cols if c in pairs.columns
and c not in ("Month", "turnaround_min")]
agg_dict = {c: "first" for c in static_feat_cols}
agg_dict["target"] = "mean" # observed bad rate
agg_dict["turnaround_min"] = "median" # median turnaround for the pair×month
agg_dict["airport_A"] = "first"
agg_dict["airport_B"] = "first"
grouped = (
pairs
.groupby(["airport_A", "airport_B", "Month", "Year"])
.agg(
n_sequences=("target", "count"),
observed_bad_rate=("target", "mean"),
median_turnaround_min=("turnaround_min", "median"),
**{c: ("target" if c == "target" else c, "first")
for c in static_feat_cols}
)
)
# Flatten multi-index columns from named agg
grouped = (
pairs
.assign(n_seq=1)
.groupby(["airport_A", "airport_B", "Month", "Year"])
.agg(
n_sequences = ("n_seq", "count"),
observed_bad_rate = ("target", "mean"),
median_turnaround_min= ("turnaround_min", "median"),
**{c: (c, "first") for c in static_feat_cols}
)
.reset_index()
)
grouped["Month"] = grouped["Month"].astype(int)
# Provisional per-year threshold (will be re-set globally in main)
threshold = grouped["observed_bad_rate"].median()
grouped["target"] = (grouped["observed_bad_rate"] > threshold).astype(int)
print(f" {year or ''}: {len(grouped):,} pair×month rows bad_rate threshold={threshold:.3f} high-risk={grouped['target'].mean():.1%}")
return grouped
# ---------------------------------------------------------------------------
# Main
# ---------------------------------------------------------------------------
def _resolve_files() -> list[str]:
"""Return best available file per year (all-carrier preferred over AA-only)."""
all_files = sorted([
os.path.join(RAW_DIR, f)
for f in os.listdir(RAW_DIR)
if f.startswith("bts_all_dfw_") and f.endswith(".parquet")
])
aa_files = sorted([
os.path.join(RAW_DIR, f)
for f in os.listdir(RAW_DIR)
if f.startswith("bts_aa_dfw_") and f.endswith(".parquet")
])
year_file = {}
for f in aa_files:
yr = f.split("_")[-1].replace(".parquet", "")
year_file[yr] = f
for f in all_files:
yr = f.split("_")[-1].replace(".parquet", "")
year_file[yr] = f
files = sorted(year_file.values())
print(f"Using files: {[os.path.basename(f) for f in files]}")
return files
def main():
files = _resolve_files()
# Build airport features by streaming one year at a time, concat the small aggregates
print("\nBuilding airport features (streaming per year)...")
ap_agg_frames = []
for fpath in files:
df_yr = pd.read_parquet(fpath)
df_yr["FlightDate"] = pd.to_datetime(df_yr["FlightDate"])
df_yr["Month"] = df_yr["FlightDate"].dt.month
df_yr["Year"] = df_yr["FlightDate"].dt.year
ap_agg_frames.append(build_airport_features(df_yr))
del df_yr
# Re-aggregate the per-year airport summaries into a single cross-year profile
ap_combined = pd.concat(ap_agg_frames, ignore_index=True)
del ap_agg_frames
airport_features = (
ap_combined.groupby(["airport", "Month"])
.mean(numeric_only=True)
.reset_index()
)
airport_features.to_parquet(os.path.join(PROCESSED_DIR, "airport_features.parquet"), index=False)
print(f"Airport features: {airport_features.shape}")
del ap_combined
# Sequences: use all-carrier files (broader training signal; airport risk is carrier-agnostic)
seq_files = sorted([
os.path.join(RAW_DIR, f)
for f in os.listdir(RAW_DIR)
if f.startswith("bts_all_dfw_") and f.endswith(".parquet")
])
print(f"\nBuilding sequences from all-carrier files: {[os.path.basename(f) for f in seq_files]}")
yearly_aggs = []
files = seq_files
for fpath in files:
year = int(os.path.basename(fpath).split("_")[-1].replace(".parquet",""))
print(f"\n--- Processing {year} ---")
df_year = pd.read_parquet(fpath)
df_year["FlightDate"] = pd.to_datetime(df_year["FlightDate"])
df_year["Month"] = df_year["FlightDate"].dt.month
df_year["Year"] = df_year["FlightDate"].dt.year
df_year["Season"] = df_year["Month"].map({
12:"winter",1:"winter",2:"winter",
3:"spring",4:"spring",5:"spring",
6:"summer",7:"summer",8:"summer",
9:"fall",10:"fall",11:"fall",
})
seqs = build_sequences(df_year)
seqs = build_feature_matrix(seqs, airport_features)
seqs = label_sequences(seqs)
# Aggregate immediately per year to avoid accumulating flight-level DFs
yearly_aggs.append(save_features(seqs, year=year))
del df_year, seqs
# Combine all per-year aggregates and re-label with a consistent threshold
all_seqs = pd.concat(yearly_aggs, ignore_index=True)
threshold = all_seqs["observed_bad_rate"].median()
all_seqs["target"] = (all_seqs["observed_bad_rate"] > threshold).astype(int)
out_path = os.path.join(PROCESSED_DIR, "sequence_features.parquet")
all_seqs.to_parquet(out_path, index=False)
print(f"\nFinal: {len(all_seqs):,} pair×month×year rows saved → {out_path}")
print(f"Global threshold: {threshold:.3f} High-risk rate: {all_seqs['target'].mean():.1%}")
if __name__ == "__main__":
main()
|