Spaces:
Sleeping
Sleeping
File size: 12,081 Bytes
be9501c | 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 | from pathlib import Path
import json
import os
import re
import uuid
import pandas as pd
from dotenv import load_dotenv
from sqlalchemy import (
Column,
DateTime,
Float,
ForeignKey,
Integer,
MetaData,
String,
Table,
create_engine,
)
from sqlalchemy import inspect
from sqlalchemy.engine import URL
from sqlalchemy.sql import func
from sqlalchemy.dialects.postgresql import JSONB
ROOT_DIR = Path(__file__).resolve().parents[1]
DATA_DIR = ROOT_DIR / "data"
ARTIFACTS_DIR = ROOT_DIR / "artifacts"
SCHEMA_PATH = ARTIFACTS_DIR / "input_schema.json"
def _load_feature_specs(schema_path: Path = SCHEMA_PATH) -> dict[str, dict]:
raw_schema = json.loads(schema_path.read_text(encoding="utf-8"))
features_raw = raw_schema.get("features", {})
if not isinstance(features_raw, dict):
return {}
normalized_features: dict[str, dict] = {}
for feature_name, spec in features_raw.items():
clean_name = str(feature_name).strip()
normalized_features[clean_name] = spec if isinstance(spec, dict) else {}
return normalized_features
FEATURE_SPECS = _load_feature_specs()
def _build_request_feature_columns() -> list[Column]:
columns: list[Column] = []
for feature_name, spec in FEATURE_SPECS.items():
feature_type = str(spec.get("type", "")).strip().lower()
if feature_type == "number":
columns.append(Column(feature_name, Float, nullable=True))
else:
columns.append(Column(feature_name, String, nullable=True))
return columns
REQUEST_FEATURE_COLUMNS = _build_request_feature_columns()
def _resolve_env_file_path() -> Path | None:
root_env = ROOT_DIR / ".env"
if root_env.exists():
return root_env
confs_dir = ROOT_DIR / "confs"
if confs_dir.exists():
env_candidates = sorted(confs_dir.rglob(".env"))
if env_candidates:
return env_candidates[0]
env_pattern_candidates = sorted(confs_dir.rglob(".env.*"))
if env_pattern_candidates:
return env_pattern_candidates[0]
return None
def _build_merged_source_dataset(data_dir: Path = DATA_DIR) -> pd.DataFrame:
eval_df = pd.read_csv(data_dir / "extrait_eval.csv")
sirh_df = pd.read_csv(data_dir / "extrait_sirh.csv")
sondage_df = pd.read_csv(data_dir / "extrait_sondage.csv")
eval_df["eval_number"] = eval_df["eval_number"].str.replace("E_", "", regex=False).astype(int)
merge_df = (
eval_df.merge(sirh_df, left_on="eval_number", right_on="id_employee", how="inner")
.merge(sondage_df, left_on="eval_number", right_on="code_sondage", how="inner")
)
merge_df.drop(columns=["eval_number", "id_employee", "code_sondage"], inplace=True)
return merge_df
def build_dataset(data_dir: Path = DATA_DIR) -> pd.DataFrame:
merge_df = _build_merged_source_dataset(data_dir=data_dir)
merge_df["augementation_salaire_precedente"] = (
merge_df["augementation_salaire_precedente"].str.replace("%", "", regex=False).astype(int)
)
col_to_drop = [
"nombre_heures_travailless",
"nombre_employee_sous_responsabilite",
"ayant_enfants",
"satisfaction_employee_environnement",
"satisfaction_employee_nature_travail",
"satisfaction_employee_equipe",
"satisfaction_employee_equilibre_pro_perso",
"note_evaluation_actuelle",
]
merge_df.drop(columns=col_to_drop, inplace=True)
salaire_moyen_par_niveau = merge_df.groupby("niveau_hierarchique_poste")["revenu_mensuel"].mean()
merge_df["salaire_moyen_niveau"] = merge_df["niveau_hierarchique_poste"].map(salaire_moyen_par_niveau)
merge_df["diff_salaire_vs_niveau"] = merge_df["revenu_mensuel"] - merge_df["salaire_moyen_niveau"]
merge_df["diff_salaire_vs_niveau_pct"] = merge_df["diff_salaire_vs_niveau"] / merge_df["salaire_moyen_niveau"]
merge_df.drop(columns=["salaire_moyen_niveau", "diff_salaire_vs_niveau"], inplace=True)
merge_df["annee_experience_totale"] = merge_df["annee_experience_totale"] + 1
merge_df["ratio_salaire_anciennete"] = merge_df["revenu_mensuel"] / merge_df["annee_experience_totale"]
col_to_drop_after = [
"annees_dans_l_entreprise",
"annee_experience_totale",
"annes_sous_responsable_actuel",
"niveau_hierarchique_poste",
]
merge_df.drop(columns=col_to_drop_after, inplace=True)
return merge_df
def build_features_engineering_variables(data_dir: Path = DATA_DIR) -> pd.DataFrame:
merge_df = _build_merged_source_dataset(data_dir=data_dir)
return merge_df[["niveau_hierarchique_poste", "annee_experience_totale"]].copy()
def get_engine_from_env():
dotenv_path = _resolve_env_file_path()
if dotenv_path is not None:
load_dotenv(dotenv_path=dotenv_path, override=False)
url = URL.create(
drivername="postgresql+psycopg2",
username=os.getenv("DB_USER"),
password=os.getenv("DB_PASSWORD"),
host=os.getenv("DB_HOST") or "127.0.0.1",
port=int(os.getenv("DB_PORT") or 5432),
database=os.getenv("DB_NAME"),
)
return create_engine(url)
# ---------------------------------------------------------------------
# API logging (requêtes + prédictions)
# ---------------------------------------------------------------------
_metadata = MetaData()
api_requests = Table(
"api_requests",
_metadata,
Column("id", String, primary_key=True),
Column("created_at", DateTime(timezone=True), server_default=func.now(), nullable=False),
Column("endpoint", String, nullable=False),
*REQUEST_FEATURE_COLUMNS,
)
api_predictions = Table(
"api_predictions",
_metadata,
Column("id", String, primary_key=True),
Column("request_id", String, ForeignKey("api_requests.id"), nullable=False),
Column("created_at", DateTime(timezone=True), server_default=func.now(), nullable=False),
Column("prediction_index", Integer, nullable=False),
Column("proba_leave", Float, nullable=False),
Column("label", Integer, nullable=False),
)
def init_api_logging_tables(engine) -> None:
"""Crée les tables de logging si elles n'existent pas."""
_metadata.create_all(engine)
def log_request_and_prediction(
engine,
*,
endpoint: str,
payload: dict,
proba_leave: list[float],
label: list[int],) -> tuple[str, str]:
request_id = str(uuid.uuid4())
if len(proba_leave) != len(label):
raise ValueError("proba_leave et label doivent avoir la même longueur")
prediction_ids: list[str] = []
request_insert_values = {
"id": request_id,
"endpoint": endpoint,
}
records = payload.get("records", []) if isinstance(payload, dict) else []
first_record = records[0] if records and isinstance(records[0], dict) else {}
for feature_name, spec in FEATURE_SPECS.items():
raw_value = first_record.get(feature_name)
if raw_value is None:
request_insert_values[feature_name] = None
continue
feature_type = str(spec.get("type", "")).strip().lower()
if feature_type == "number":
request_insert_values[feature_name] = float(raw_value)
else:
request_insert_values[feature_name] = str(raw_value)
with engine.begin() as conn:
conn.execute(
api_requests.insert().values(
**request_insert_values,
)
)
for idx, (proba_value, label_value) in enumerate(zip(proba_leave, label)):
prediction_id = str(uuid.uuid4())
prediction_ids.append(prediction_id)
conn.execute(
api_predictions.insert().values(
id=prediction_id,
request_id=request_id,
prediction_index=idx,
proba_leave=float(proba_value),
label=int(label_value),
)
)
first_prediction_id = prediction_ids[0] if prediction_ids else ""
return request_id, first_prediction_id
def full_dataset_to_bdd(data_dir: Path = DATA_DIR, table_name: str = "dataset_final") -> pd.DataFrame:
merge_df = build_dataset(data_dir=data_dir)
engine = get_engine_from_env()
merge_df.to_sql(name=table_name, con=engine, if_exists="replace", index=False)
print(f"Dataset envoyé avec succès dans la table '{table_name}'")
return merge_df
def features_engineering_variables_to_bdd(
data_dir: Path = DATA_DIR,
table_name: str = "features_engineering_variable",
) -> pd.DataFrame:
features_df = build_features_engineering_variables(data_dir=data_dir)
engine = get_engine_from_env()
features_df.to_sql(name=table_name, con=engine, if_exists="replace", index=False)
print(f"Features engineering envoyées avec succès dans la table '{table_name}'")
return features_df
def init_feature_tables_if_missing(
engine,
data_dir: Path = DATA_DIR,
dataset_table_name: str = "dataset_final",
features_table_name: str = "features_engineering_variable",
) -> None:
inspector = inspect(engine)
if not inspector.has_table(dataset_table_name):
build_dataset(data_dir=data_dir).to_sql(
name=dataset_table_name,
con=engine,
if_exists="replace",
index=False,
)
if not inspector.has_table(features_table_name):
build_features_engineering_variables(data_dir=data_dir).to_sql(
name=features_table_name,
con=engine,
if_exists="replace",
index=False,
)
def build_payload_from_bdd_row(
row_number: int,
table_name: str = "dataset_final",
features_table_name: str = "features_engineering_variable",
engine=None,
) -> dict:
if row_number < 1:
raise ValueError("row_number doit être >= 1")
if not re.fullmatch(r"[A-Za-z0-9_]+", table_name):
raise ValueError("table_name invalide")
if not re.fullmatch(r"[A-Za-z0-9_]+", features_table_name):
raise ValueError("features_table_name invalide")
db_engine = engine if engine is not None else get_engine_from_env()
offset = row_number - 1
dataset_query = f'SELECT * FROM "{table_name}" LIMIT 1 OFFSET {offset}'
features_query = f'SELECT * FROM "{features_table_name}" LIMIT 1 OFFSET {offset}'
dataset_row_df = pd.read_sql_query(dataset_query, con=db_engine)
features_row_df = pd.read_sql_query(features_query, con=db_engine)
if dataset_row_df.empty:
raise IndexError(
f"Aucune ligne trouvée pour row_number={row_number} dans la table '{table_name}'"
)
if features_row_df.empty:
raise IndexError(
f"Aucune ligne trouvée pour row_number={row_number} dans la table '{features_table_name}'"
)
dataset_row_df = dataset_row_df.iloc[:, :-2]
dataset_row = dataset_row_df.iloc[0]
features_row = features_row_df.iloc[0]
record: dict[str, float | str | None] = {}
required_feature_columns = ["niveau_hierarchique_poste", "annee_experience_totale"]
for col in required_feature_columns:
if col not in features_row.index:
raise KeyError(
f"La colonne '{col}' est absente de la table '{features_table_name}'"
)
for feature_name, spec in FEATURE_SPECS.items():
if feature_name in dataset_row.index:
value = dataset_row[feature_name]
elif feature_name in required_feature_columns:
value = features_row[feature_name]
else:
raise KeyError(
f"La colonne '{feature_name}' est absente des tables '{table_name}' et '{features_table_name}'"
)
if pd.isna(value):
record[feature_name] = None
continue
feature_type = str(spec.get("type", "")).strip().lower()
if feature_type == "number":
record[feature_name] = float(value)
else:
record[feature_name] = str(value)
return {"records": [record]}
|