Spaces:
Sleeping
Sleeping
File size: 12,780 Bytes
1d0150f | 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 | from __future__ import annotations
import os
import sys
import logging
from dataclasses import dataclass
import pandas as pd
from sqlalchemy import create_engine, text
from dotenv import load_dotenv, find_dotenv
# CONFIG & LOGGING
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s | %(levelname)s | %(message)s",
)
logger = logging.getLogger("technova.etl")
# SETTINGS
@dataclass(frozen=True)
class Settings:
# RAW schema tables
raw_employees: str = "raw.employees"
raw_snapshots: str = "raw.employee_snapshots"
raw_surveys: str = "raw.surveys"
ground_truth: str = "raw.ground_truth"
# CLEAN destination
dst_schema: str = "clean"
dst_name: str = "ml_features_employees"
dst_qualified: str = "clean.ml_features_employees"
# UTILS
def get_engine():
load_dotenv(find_dotenv())
db_url = os.getenv("DATABASE_URL")
if not db_url:
raise RuntimeError("DATABASE_URL manquant dans .env")
return create_engine(db_url, pool_pre_ping=True)
# TRANSFORMS
class Transform:
@staticmethod
def clean_raw_inputs(df: pd.DataFrame) -> pd.DataFrame:
df = df.copy()
# % -> float
if "augmentation_salaire_precedente" in df.columns:
s = df["augmentation_salaire_precedente"].astype("string")
s = s.str.replace("%", "", regex=False)
df["augmentation_salaire_precedente"] = pd.to_numeric(s, errors="coerce")
# Oui/Non -> 1/0
if "heures_supplementaires" in df.columns:
df["heures_supplementaires"] = df["heures_supplementaires"].map(
{"Oui": 1, "Non": 0, "oui": 1, "non": 0, 1: 1, 0: 0, True: 1, False: 0}
)
# Genre -> 1/0
if "genre" in df.columns:
df["genre"] = df["genre"].map({"M": 1, "F": 0, 1: 1, 0: 0})
# Fréquence déplacement -> 0/1/2 (+ éventuellement 3 si tu veux)
if "frequence_deplacement" in df.columns:
df["frequence_deplacement"] = df["frequence_deplacement"].map(
{
"Aucun": 0,
"Occasionnel": 1,
"Frequent": 2,
"Fréquent": 2,
0: 0, 1: 1, 2: 2, 3: 3,
}
)
numeric_cols = [
"employee_id",
"age",
"revenu_mensuel",
"niveau_education",
"distance_domicile_travail",
"note_evaluation_precedente",
"note_evaluation_actuelle",
"niveau_hierarchique_poste",
"nombre_experiences_precedentes",
"annee_experience_totale",
"annees_dans_l_entreprise",
"annees_dans_le_poste_actuel",
"annees_depuis_la_derniere_promotion",
"annees_sous_responsable_actuel",
"nombre_participation_pee",
"nb_formations_suivies",
"a_quitte_l_entreprise"
]
for c in numeric_cols:
if c in df.columns:
df[c] = pd.to_numeric(df[c], errors="coerce")
return df
@staticmethod
def feature_engineering(df: pd.DataFrame) -> pd.DataFrame:
df = df.copy()
sat_cols = [
"satisfaction_employee_environnement",
"satisfaction_employee_nature_travail",
"satisfaction_employee_equipe",
"satisfaction_employee_equilibre_pro_perso"
]
missing = [c for c in sat_cols if c not in df.columns]
if missing:
raise KeyError(f"Colonnes satisfaction manquantes (raw.surveys?) : {missing}")
# satisfaction moyenne
df["satisfaction_moyenne"] = df[sat_cols].mean(axis=1)
# features dérivées
df["nonlineaire_participation_pee"] = (
df["nombre_participation_pee"]
/ (df["nombre_participation_pee"] + df["annees_dans_l_entreprise"] + 1)
)
df["ratio_heures_sup_salaire"] = (
df["heures_supplementaires"] / (df["revenu_mensuel"] + 1)
)
d = df["distance_domicile_travail"]
df["nonlinaire_charge_contrainte"] = (
df["heures_supplementaires"] * d / (d + 10) / (d + 10)
)
df["nonlinaire_surmenage_insatisfaction"] = (
df["heures_supplementaires"] * (1 - df["satisfaction_moyenne"])
)
df["jeune_surcharge"] = (
(df["age"] < 30) & (df["heures_supplementaires"] == 1)
).astype(int)
df["anciennete_sans_promotion"] = (
(df["annees_dans_l_entreprise"] - df["annees_depuis_la_derniere_promotion"])
/ (df["annees_dans_l_entreprise"] + 1)
)
df["mobilite_carriere"] = (
df["nombre_experiences_precedentes"]
/ (df["annee_experience_totale"] + 1)
)
df["risque_global"] = (
df["ratio_heures_sup_salaire"]
* df["anciennete_sans_promotion"]
* (1 - df["satisfaction_moyenne"])
)
return df
@staticmethod
def suppression_features(df: pd.DataFrame) -> pd.DataFrame:
return df.drop(
columns=[
"satisfaction_employee_environnement",
"satisfaction_employee_nature_travail",
"satisfaction_employee_equipe",
"satisfaction_employee_equilibre_pro_perso"
],
errors="ignore",
)
# DESTINATION COLUMNS
DEST_COLS = [
"employee_id",
"note_evaluation_precedente",
"niveau_hierarchique_poste",
"note_evaluation_actuelle",
"heures_supplementaires",
"augmentation_salaire_precedente",
"age",
"genre",
"revenu_mensuel",
"statut_marital",
"departement",
"poste",
"nombre_experiences_precedentes",
"annee_experience_totale",
"annees_dans_l_entreprise",
"annees_dans_le_poste_actuel",
"a_quitte_l_entreprise",
"nombre_participation_pee",
"nb_formations_suivies",
"distance_domicile_travail",
"niveau_education",
"domaine_etude",
"frequence_deplacement",
"annees_depuis_la_derniere_promotion",
"annees_sous_responsable_actuel",
"satisfaction_moyenne",
"nonlineaire_participation_pee",
"ratio_heures_sup_salaire",
"nonlinaire_charge_contrainte",
"nonlinaire_surmenage_insatisfaction",
"jeune_surcharge",
"anciennete_sans_promotion",
"mobilite_carriere",
"risque_global"
]
def build_sql_master(s: Settings) -> str:
return f"""
WITH last_snapshot AS (
SELECT DISTINCT ON (employee_id)
employee_id,
nombre_experiences_precedentes,
annee_experience_totale,
annees_dans_l_entreprise,
annees_dans_le_poste_actuel,
annees_sous_responsable_actuel,
niveau_hierarchique_poste,
revenu_mensuel,
augmentation_salaire_precedente,
heures_supplementaires,
nombre_participation_pee,
nb_formations_suivies,
frequence_deplacement,
annees_depuis_la_derniere_promotion,
created_at
FROM {s.raw_snapshots}
ORDER BY employee_id, created_at DESC
),
last_survey AS (
SELECT DISTINCT ON (employee_id)
employee_id,
note_evaluation_precedente,
note_evaluation_actuelle,
satisfaction_employee_environnement,
satisfaction_employee_nature_travail,
satisfaction_employee_equipe,
satisfaction_employee_equilibre_pro_perso,
created_at
FROM {s.raw_surveys}
ORDER BY employee_id, created_at DESC
),
last_truth AS (
SELECT DISTINCT ON (employee_id)
employee_id,
a_quitte_l_entreprise,
date_event
FROM {s.ground_truth}
ORDER BY employee_id, date_event DESC
)
SELECT
e.id AS employee_id,
e.age,
e.genre,
e.statut_marital,
e.niveau_education,
e.domaine_etude,
e.departement,
e.poste,
e.distance_domicile_travail,
s.nombre_experiences_precedentes,
s.annee_experience_totale,
s.annees_dans_l_entreprise,
s.annees_dans_le_poste_actuel,
s.annees_sous_responsable_actuel,
s.niveau_hierarchique_poste,
s.revenu_mensuel,
s.augmentation_salaire_precedente,
s.heures_supplementaires,
s.nombre_participation_pee,
s.nb_formations_suivies,
s.frequence_deplacement,
s.annees_depuis_la_derniere_promotion,
sv.note_evaluation_precedente,
sv.note_evaluation_actuelle,
sv.satisfaction_employee_environnement,
sv.satisfaction_employee_nature_travail,
sv.satisfaction_employee_equipe,
sv.satisfaction_employee_equilibre_pro_perso,
COALESCE(gt.a_quitte_l_entreprise, 0) AS a_quitte_l_entreprise
FROM {s.raw_employees} e
LEFT JOIN last_snapshot s ON s.employee_id = e.id
LEFT JOIN last_survey sv ON sv.employee_id = e.id
LEFT JOIN last_truth gt ON gt.employee_id = e.id
WHERE s.employee_id IS NOT NULL
AND sv.employee_id IS NOT NULL
;
"""
# ETL STEPS
def fetch_master_df(engine, sql: str) -> pd.DataFrame:
with engine.connect() as conn:
return pd.read_sql(text(sql), conn)
def validate_columns(df: pd.DataFrame, dst_qualified: str) -> None:
missing = [c for c in DEST_COLS if c not in df.columns]
if missing:
raise KeyError(f"Colonnes manquantes pour insertion dans {dst_qualified}: {missing}")
def validate_quality(df: pd.DataFrame) -> None:
critical = ["employee_id", "age", "revenu_mensuel", "heures_supplementaires", "a_quitte_l_entreprise"]
bad = [c for c in critical if c in df.columns and df[c].isna().mean() > 0.20]
if bad:
raise ValueError(f"Trop de NaN sur colonnes critiques (>20%): {bad}")
def truncate_destination(engine, dst_qualified: str) -> None:
with engine.begin() as conn:
conn.execute(text(f"TRUNCATE TABLE {dst_qualified} RESTART IDENTITY;"))
def enforce_not_null_ready(df: pd.DataFrame) -> pd.DataFrame:
out = df.copy()
out = out.dropna(subset=DEST_COLS)
# cast ints propres (pandas peut garder float si NaN existe, mais là on a drop)
int_like = [
"employee_id",
"note_evaluation_precedente",
"niveau_hierarchique_poste",
"note_evaluation_actuelle",
"heures_supplementaires",
"age",
"genre",
"revenu_mensuel",
"nombre_experiences_precedentes",
"annee_experience_totale",
"annees_dans_l_entreprise",
"annees_dans_le_poste_actuel",
"a_quitte_l_entreprise",
"nombre_participation_pee",
"nb_formations_suivies",
"distance_domicile_travail",
"niveau_education",
"frequence_deplacement",
"annees_depuis_la_derniere_promotion",
"annees_sous_responsable_actuel",
"jeune_surcharge"
]
for c in int_like:
if c in out.columns:
out[c] = out[c].astype(int)
return out
def insert_destination(engine, df: pd.DataFrame, dst_schema: str, dst_name: str, chunk_size: int = 2000) -> int:
df_out = df[DEST_COLS].copy()
with engine.begin() as conn:
df_out.to_sql(
name=dst_name,
schema=dst_schema,
con=conn,
if_exists="append",
index=False,
method="multi",
chunksize=chunk_size,
)
return len(df_out)
# MAIN
def main():
refresh = "--refresh" in sys.argv
dry_run = "--dry-run" in sys.argv
s = Settings()
engine = get_engine()
sql_master = build_sql_master(s)
logger.info("Build master DF depuis raw tables")
df = fetch_master_df(engine, sql_master)
logger.info("Master DF: %s lignes | %s colonnes", df.shape[0], df.shape[1])
logger.info("Nettoyage types / mapping")
df = Transform.clean_raw_inputs(df)
logger.info("Feature engineering")
df = Transform.feature_engineering(df)
logger.info("Drop colonnes intermédiaires")
df = Transform.suppression_features(df)
logger.info("Validation colonnes destination")
validate_columns(df, s.dst_qualified)
logger.info("Contrôle qualité (NaN critiques)")
validate_quality(df)
logger.info("Préparation NOT NULL (drop rows invalides + cast ints)")
before = len(df)
df = enforce_not_null_ready(df)
after = len(df)
if after < before:
logger.warning("Lignes retirées (NULL sur colonnes clean NOT NULL): %s -> %s", before, after)
if dry_run:
logger.info("DRY-RUN: aucune écriture en base. Aperçu:")
logger.info("Colonnes finales: %s", list(df[DEST_COLS].columns))
logger.info("Head:\n%s", df[DEST_COLS].head(5).to_string(index=False))
return
if refresh:
logger.info("Mode REFRESH: TRUNCATE %s", s.dst_qualified)
truncate_destination(engine, s.dst_qualified)
logger.info("Insertion vers %s", s.dst_qualified)
n = insert_destination(engine, df, s.dst_schema, s.dst_name)
logger.info("OK: %s lignes insérées dans %s", n, s.dst_qualified)
if __name__ == "__main__":
main()
|