Diaure commited on
Commit
3255f93
·
verified ·
1 Parent(s): 648946d

CD: update from GitHub main

Browse files
.gitattributes CHANGED
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ Images/CM_modele_retenu.png filter=lfs diff=lfs merge=lfs -text
37
+ Images/SHAP_global.png filter=lfs diff=lfs merge=lfs -text
38
+ Images/SHAP_local.png filter=lfs diff=lfs merge=lfs -text
.gitignore ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Donnees_Brutes/
2
+ *.csv
3
+ data/
4
+ mlruns/
5
+ mlflow.db
6
+ __pycache__/
7
+ *.pkl
8
+ *.onnx
9
+ *.h5
10
+ **/artifacts/
11
+ artifacts/
12
+ artefacts/
13
+ *.pkl
14
+ *.ipynb_checkpoints/
App/__init__.py ADDED
File without changes
App/app.py ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ import joblib
3
+ import pandas as pd
4
+ from models import ClientFeatures
5
+ import sys, os
6
+
7
+ # ROOT = os.path.dirname(os.path.abspath(__file__))
8
+ # sys.path.append(ROOT)
9
+
10
+ # Charger le pipeline
11
+ pipe = joblib.load("./BestModel/pipeline_complet.joblib")
12
+
13
+ # Charger le seuil optimal
14
+ threshold = joblib.load("./BestModel/best_threshold.joblib")
15
+
16
+ # Récupérer l'exemple généré automatiquement dans models.py
17
+ example = ClientFeatures.model_config["json_schema_extra"]["example"]
18
+
19
+
20
+ def predict_gradio(*args):
21
+ try:
22
+ # Reconstruction automatique
23
+ data_dict = {name: value for (name, value) in zip(ClientFeatures.model_fields.keys(), args)}
24
+ df = pd.DataFrame([data_dict])
25
+
26
+ # Conversion robuste : transforme tout ce qui peut être un nombre
27
+ df = df.apply(lambda col: pd.to_numeric(col, errors="coerce"))
28
+
29
+ # Validation métier
30
+ if "DAYS_BIRTH" in df.columns:
31
+ raw_age = df["DAYS_BIRTH"].iloc[0]
32
+
33
+ # Forcer la conversion en float de manière robuste
34
+ try:
35
+ age = float(raw_age)
36
+ except (TypeError, ValueError):
37
+ age = None
38
+
39
+ # Si on a bien un nombre, on applique la règle métier
40
+ if age is not None and (age >= 0 or age > -18 * 365):
41
+ return "Âge invalide", None, None
42
+
43
+
44
+ if "AMT_INCOME_TOTAL" in df.columns:
45
+ income = df["AMT_INCOME_TOTAL"].iloc[0]
46
+ if pd.notna(income) and income <= 0:
47
+ return "Revenu invalide", None, None
48
+
49
+ # Prédiction
50
+ score = pipe.predict_proba(df)[0][1]
51
+ decision = "ACCORDÉ" if score < threshold else "REFUSÉ"
52
+
53
+ return decision, float(score), float(threshold)
54
+
55
+ except Exception as e:
56
+ return f"ERREUR : {str(e)}", None, None
57
+
58
+
59
+
60
+ # Génération automatique des inputs Gradio AVEC VALEURS PAR DÉFAUT
61
+ inputs = []
62
+ for name, field in ClientFeatures.model_fields.items():
63
+
64
+ default_value = example.get(name, None)
65
+
66
+ if field.annotation == float:
67
+ inputs.append(gr.Number(label=name, value=default_value))
68
+ elif field.annotation == int:
69
+ inputs.append(gr.Number(label=name, value=default_value))
70
+ elif field.annotation == bool:
71
+ inputs.append(gr.Checkbox(label=name, value=default_value))
72
+ elif field.annotation == str:
73
+ inputs.append(gr.Textbox(label=name, value=default_value))
74
+ else:
75
+ inputs.append(gr.Textbox(label=name, value=default_value))
76
+
77
+
78
+ # Sorties Gradio
79
+ outputs = [
80
+ gr.Textbox(label="Décision"),
81
+ gr.Number(label="Score"),
82
+ gr.Number(label="Seuil Optimal"),
83
+ ]
84
+
85
+ # Interface Gradio
86
+ app = gr.Interface(
87
+ fn=predict_gradio,
88
+ inputs=inputs,
89
+ outputs=outputs,
90
+ title="Scoring Crédit - Interface Gradio",
91
+ description="Interface interactive pour tester le modèle de scoring crédit."
92
+ )
93
+
94
+ app.launch()
App/column_mapping.json ADDED
@@ -0,0 +1,329 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "SK_ID_CURR": "SK_ID_CURR",
3
+ "NAME_CONTRACT_TYPE": "NAME_CONTRACT_TYPE",
4
+ "CODE_GENDER": "CODE_GENDER",
5
+ "FLAG_OWN_CAR": "FLAG_OWN_CAR",
6
+ "FLAG_OWN_REALTY": "FLAG_OWN_REALTY",
7
+ "CNT_CHILDREN": "CNT_CHILDREN",
8
+ "AMT_INCOME_TOTAL": "AMT_INCOME_TOTAL",
9
+ "AMT_GOODS_PRICE": "AMT_GOODS_PRICE",
10
+ "NAME_TYPE_SUITE": "NAME_TYPE_SUITE",
11
+ "NAME_INCOME_TYPE": "NAME_INCOME_TYPE",
12
+ "NAME_EDUCATION_TYPE": "NAME_EDUCATION_TYPE",
13
+ "NAME_FAMILY_STATUS": "NAME_FAMILY_STATUS",
14
+ "NAME_HOUSING_TYPE": "NAME_HOUSING_TYPE",
15
+ "REGION_POPULATION_RELATIVE": "REGION_POPULATION_RELATIVE",
16
+ "DAYS_BIRTH": "DAYS_BIRTH",
17
+ "DAYS_EMPLOYED": "DAYS_EMPLOYED",
18
+ "DAYS_REGISTRATION": "DAYS_REGISTRATION",
19
+ "DAYS_ID_PUBLISH": "DAYS_ID_PUBLISH",
20
+ "OWN_CAR_AGE": "OWN_CAR_AGE",
21
+ "FLAG_MOBIL": "FLAG_MOBIL",
22
+ "FLAG_EMP_PHONE": "FLAG_EMP_PHONE",
23
+ "FLAG_WORK_PHONE": "FLAG_WORK_PHONE",
24
+ "FLAG_CONT_MOBILE": "FLAG_CONT_MOBILE",
25
+ "FLAG_PHONE": "FLAG_PHONE",
26
+ "FLAG_EMAIL": "FLAG_EMAIL",
27
+ "OCCUPATION_TYPE": "OCCUPATION_TYPE",
28
+ "REGION_RATING_CLIENT_W_CITY": "REGION_RATING_CLIENT_W_CITY",
29
+ "WEEKDAY_APPR_PROCESS_START": "WEEKDAY_APPR_PROCESS_START",
30
+ "HOUR_APPR_PROCESS_START": "HOUR_APPR_PROCESS_START",
31
+ "REG_REGION_NOT_LIVE_REGION": "REG_REGION_NOT_LIVE_REGION",
32
+ "REG_REGION_NOT_WORK_REGION": "REG_REGION_NOT_WORK_REGION",
33
+ "LIVE_REGION_NOT_WORK_REGION": "LIVE_REGION_NOT_WORK_REGION",
34
+ "REG_CITY_NOT_LIVE_CITY": "REG_CITY_NOT_LIVE_CITY",
35
+ "REG_CITY_NOT_WORK_CITY": "REG_CITY_NOT_WORK_CITY",
36
+ "LIVE_CITY_NOT_WORK_CITY": "LIVE_CITY_NOT_WORK_CITY",
37
+ "ORGANIZATION_TYPE": "ORGANIZATION_TYPE",
38
+ "EXT_SOURCE_1": "EXT_SOURCE_1",
39
+ "EXT_SOURCE_2": "EXT_SOURCE_2",
40
+ "EXT_SOURCE_3": "EXT_SOURCE_3",
41
+ "BASEMENTAREA_AVG": "BASEMENTAREA_AVG",
42
+ "ELEVATORS_AVG": "ELEVATORS_AVG",
43
+ "ENTRANCES_AVG": "ENTRANCES_AVG",
44
+ "FLOORSMAX_AVG": "FLOORSMAX_AVG",
45
+ "NONLIVINGAPARTMENTS_AVG": "NONLIVINGAPARTMENTS_AVG",
46
+ "NONLIVINGAREA_AVG": "NONLIVINGAREA_AVG",
47
+ "YEARS_BEGINEXPLUATATION_MEDI": "YEARS_BEGINEXPLUATATION_MEDI",
48
+ "YEARS_BUILD_MEDI": "YEARS_BUILD_MEDI",
49
+ "COMMONAREA_MEDI": "COMMONAREA_MEDI",
50
+ "LANDAREA_MEDI": "LANDAREA_MEDI",
51
+ "FONDKAPREMONT_MODE": "FONDKAPREMONT_MODE",
52
+ "HOUSETYPE_MODE": "HOUSETYPE_MODE",
53
+ "WALLSMATERIAL_MODE": "WALLSMATERIAL_MODE",
54
+ "EMERGENCYSTATE_MODE": "EMERGENCYSTATE_MODE",
55
+ "OBS_30_CNT_SOCIAL_CIRCLE": "OBS_30_CNT_SOCIAL_CIRCLE",
56
+ "DEF_30_CNT_SOCIAL_CIRCLE": "DEF_30_CNT_SOCIAL_CIRCLE",
57
+ "DAYS_LAST_PHONE_CHANGE": "DAYS_LAST_PHONE_CHANGE",
58
+ "FLAG_DOCUMENT_2": "FLAG_DOCUMENT_2",
59
+ "FLAG_DOCUMENT_3": "FLAG_DOCUMENT_3",
60
+ "FLAG_DOCUMENT_4": "FLAG_DOCUMENT_4",
61
+ "FLAG_DOCUMENT_5": "FLAG_DOCUMENT_5",
62
+ "FLAG_DOCUMENT_6": "FLAG_DOCUMENT_6",
63
+ "FLAG_DOCUMENT_7": "FLAG_DOCUMENT_7",
64
+ "FLAG_DOCUMENT_8": "FLAG_DOCUMENT_8",
65
+ "FLAG_DOCUMENT_9": "FLAG_DOCUMENT_9",
66
+ "FLAG_DOCUMENT_10": "FLAG_DOCUMENT_10",
67
+ "FLAG_DOCUMENT_11": "FLAG_DOCUMENT_11",
68
+ "FLAG_DOCUMENT_12": "FLAG_DOCUMENT_12",
69
+ "FLAG_DOCUMENT_13": "FLAG_DOCUMENT_13",
70
+ "FLAG_DOCUMENT_14": "FLAG_DOCUMENT_14",
71
+ "FLAG_DOCUMENT_15": "FLAG_DOCUMENT_15",
72
+ "FLAG_DOCUMENT_16": "FLAG_DOCUMENT_16",
73
+ "FLAG_DOCUMENT_17": "FLAG_DOCUMENT_17",
74
+ "FLAG_DOCUMENT_18": "FLAG_DOCUMENT_18",
75
+ "FLAG_DOCUMENT_19": "FLAG_DOCUMENT_19",
76
+ "FLAG_DOCUMENT_20": "FLAG_DOCUMENT_20",
77
+ "FLAG_DOCUMENT_21": "FLAG_DOCUMENT_21",
78
+ "AMT_REQ_CREDIT_BUREAU_HOUR": "AMT_REQ_CREDIT_BUREAU_HOUR",
79
+ "AMT_REQ_CREDIT_BUREAU_DAY": "AMT_REQ_CREDIT_BUREAU_DAY",
80
+ "AMT_REQ_CREDIT_BUREAU_WEEK": "AMT_REQ_CREDIT_BUREAU_WEEK",
81
+ "AMT_REQ_CREDIT_BUREAU_MON": "AMT_REQ_CREDIT_BUREAU_MON",
82
+ "AMT_REQ_CREDIT_BUREAU_QRT": "AMT_REQ_CREDIT_BUREAU_QRT",
83
+ "BUREAU_SK_ID_BUREAU_max": "BUREAU_SK_ID_BUREAU_max",
84
+ "BUREAU_DAYS_CREDIT_mean": "BUREAU_DAYS_CREDIT_mean",
85
+ "BUREAU_DAYS_CREDIT_max": "BUREAU_DAYS_CREDIT_max",
86
+ "BUREAU_DAYS_CREDIT_min": "BUREAU_DAYS_CREDIT_min",
87
+ "BUREAU_CREDIT_DAY_OVERDUE_mean": "BUREAU_CREDIT_DAY_OVERDUE_mean",
88
+ "BUREAU_CREDIT_DAY_OVERDUE_sum": "BUREAU_CREDIT_DAY_OVERDUE_sum",
89
+ "BUREAU_DAYS_CREDIT_ENDDATE_mean": "BUREAU_DAYS_CREDIT_ENDDATE_mean",
90
+ "BUREAU_DAYS_CREDIT_ENDDATE_max": "BUREAU_DAYS_CREDIT_ENDDATE_max",
91
+ "BUREAU_DAYS_CREDIT_ENDDATE_min": "BUREAU_DAYS_CREDIT_ENDDATE_min",
92
+ "BUREAU_DAYS_CREDIT_ENDDATE_sum": "BUREAU_DAYS_CREDIT_ENDDATE_sum",
93
+ "BUREAU_DAYS_ENDDATE_FACT_max": "BUREAU_DAYS_ENDDATE_FACT_max",
94
+ "BUREAU_DAYS_ENDDATE_FACT_sum": "BUREAU_DAYS_ENDDATE_FACT_sum",
95
+ "BUREAU_AMT_CREDIT_MAX_OVERDUE_max": "BUREAU_AMT_CREDIT_MAX_OVERDUE_max",
96
+ "BUREAU_CNT_CREDIT_PROLONG_max": "BUREAU_CNT_CREDIT_PROLONG_max",
97
+ "BUREAU_CNT_CREDIT_PROLONG_sum": "BUREAU_CNT_CREDIT_PROLONG_sum",
98
+ "BUREAU_AMT_CREDIT_SUM_mean": "BUREAU_AMT_CREDIT_SUM_mean",
99
+ "BUREAU_AMT_CREDIT_SUM_min": "BUREAU_AMT_CREDIT_SUM_min",
100
+ "BUREAU_AMT_CREDIT_SUM_sum": "BUREAU_AMT_CREDIT_SUM_sum",
101
+ "BUREAU_AMT_CREDIT_SUM_DEBT_max": "BUREAU_AMT_CREDIT_SUM_DEBT_max",
102
+ "BUREAU_AMT_CREDIT_SUM_DEBT_sum": "BUREAU_AMT_CREDIT_SUM_DEBT_sum",
103
+ "BUREAU_AMT_CREDIT_SUM_LIMIT_mean": "BUREAU_AMT_CREDIT_SUM_LIMIT_mean",
104
+ "BUREAU_AMT_CREDIT_SUM_LIMIT_sum": "BUREAU_AMT_CREDIT_SUM_LIMIT_sum",
105
+ "BUREAU_AMT_CREDIT_SUM_OVERDUE_max": "BUREAU_AMT_CREDIT_SUM_OVERDUE_max",
106
+ "BUREAU_AMT_CREDIT_SUM_OVERDUE_sum": "BUREAU_AMT_CREDIT_SUM_OVERDUE_sum",
107
+ "BUREAU_AMT_ANNUITY_mean": "BUREAU_AMT_ANNUITY_mean",
108
+ "BUREAU_AMT_ANNUITY_min": "BUREAU_AMT_ANNUITY_min",
109
+ "BUREAU_MONTHS_BALANCE_max": "BUREAU_MONTHS_BALANCE_max",
110
+ "BUREAU_CREDIT_ACTIVE_NUNIQUE": "BUREAU_CREDIT_ACTIVE_NUNIQUE",
111
+ "BUREAU_CREDIT_CURRENCY_NUNIQUE": "BUREAU_CREDIT_CURRENCY_NUNIQUE",
112
+ "BUREAU_CREDIT_TYPE_NUNIQUE": "BUREAU_CREDIT_TYPE_NUNIQUE",
113
+ "BUREAU_STATUS_NUNIQUE": "BUREAU_STATUS_NUNIQUE",
114
+ "BUREAU_CREDIT_ACTIVE_Bad_debt": "BUREAU_CREDIT_ACTIVE_Bad debt",
115
+ "BUREAU_CREDIT_ACTIVE_Sold": "BUREAU_CREDIT_ACTIVE_Sold",
116
+ "BUREAU_CREDIT_CURRENCY_currency_2": "BUREAU_CREDIT_CURRENCY_currency 2",
117
+ "BUREAU_CREDIT_CURRENCY_currency_3": "BUREAU_CREDIT_CURRENCY_currency 3",
118
+ "BUREAU_CREDIT_CURRENCY_currency_4": "BUREAU_CREDIT_CURRENCY_currency 4",
119
+ "BUREAU_CREDIT_TYPE_Another_type_of_loan": "BUREAU_CREDIT_TYPE_Another type of loan",
120
+ "BUREAU_CREDIT_TYPE_Car_loan": "BUREAU_CREDIT_TYPE_Car loan",
121
+ "BUREAU_CREDIT_TYPE_Cash_loan__non_earmarked_": "BUREAU_CREDIT_TYPE_Cash loan (non-earmarked)",
122
+ "BUREAU_CREDIT_TYPE_Interbank_credit": "BUREAU_CREDIT_TYPE_Interbank credit",
123
+ "BUREAU_CREDIT_TYPE_Loan_for_business_development": "BUREAU_CREDIT_TYPE_Loan for business development",
124
+ "BUREAU_CREDIT_TYPE_Loan_for_purchase_of_shares__margin_lending_": "BUREAU_CREDIT_TYPE_Loan for purchase of shares (margin lending)",
125
+ "BUREAU_CREDIT_TYPE_Loan_for_the_purchase_of_equipment": "BUREAU_CREDIT_TYPE_Loan for the purchase of equipment",
126
+ "BUREAU_CREDIT_TYPE_Loan_for_working_capital_replenishment": "BUREAU_CREDIT_TYPE_Loan for working capital replenishment",
127
+ "BUREAU_CREDIT_TYPE_Microloan": "BUREAU_CREDIT_TYPE_Microloan",
128
+ "BUREAU_CREDIT_TYPE_Mobile_operator_loan": "BUREAU_CREDIT_TYPE_Mobile operator loan",
129
+ "BUREAU_CREDIT_TYPE_Mortgage": "BUREAU_CREDIT_TYPE_Mortgage",
130
+ "BUREAU_CREDIT_TYPE_Real_estate_loan": "BUREAU_CREDIT_TYPE_Real estate loan",
131
+ "BUREAU_CREDIT_TYPE_Unknown_type_of_loan": "BUREAU_CREDIT_TYPE_Unknown type of loan",
132
+ "BUREAU_STATUS_1": "BUREAU_STATUS_1",
133
+ "BUREAU_STATUS_2": "BUREAU_STATUS_2",
134
+ "BUREAU_STATUS_4": "BUREAU_STATUS_4",
135
+ "BUREAU_STATUS_5": "BUREAU_STATUS_5",
136
+ "PREV_AMT_ANNUITY_mean": "PREV_AMT_ANNUITY_mean",
137
+ "PREV_AMT_ANNUITY_min": "PREV_AMT_ANNUITY_min",
138
+ "PREV_AMT_APPLICATION_min": "PREV_AMT_APPLICATION_min",
139
+ "PREV_AMT_DOWN_PAYMENT_sum": "PREV_AMT_DOWN_PAYMENT_sum",
140
+ "PREV_HOUR_APPR_PROCESS_START_mean": "PREV_HOUR_APPR_PROCESS_START_mean",
141
+ "PREV_NFLAG_LAST_APPL_IN_DAY_max": "PREV_NFLAG_LAST_APPL_IN_DAY_max",
142
+ "PREV_NFLAG_LAST_APPL_IN_DAY_min": "PREV_NFLAG_LAST_APPL_IN_DAY_min",
143
+ "PREV_RATE_DOWN_PAYMENT_max": "PREV_RATE_DOWN_PAYMENT_max",
144
+ "PREV_RATE_DOWN_PAYMENT_sum": "PREV_RATE_DOWN_PAYMENT_sum",
145
+ "PREV_RATE_INTEREST_PRIMARY_max": "PREV_RATE_INTEREST_PRIMARY_max",
146
+ "PREV_RATE_INTEREST_PRIMARY_sum": "PREV_RATE_INTEREST_PRIMARY_sum",
147
+ "PREV_RATE_INTEREST_PRIVILEGED_max": "PREV_RATE_INTEREST_PRIVILEGED_max",
148
+ "PREV_SELLERPLACE_AREA_sum": "PREV_SELLERPLACE_AREA_sum",
149
+ "PREV_CNT_PAYMENT_max": "PREV_CNT_PAYMENT_max",
150
+ "PREV_CNT_PAYMENT_min": "PREV_CNT_PAYMENT_min",
151
+ "PREV_DAYS_FIRST_DRAWING_mean": "PREV_DAYS_FIRST_DRAWING_mean",
152
+ "PREV_DAYS_FIRST_DRAWING_max": "PREV_DAYS_FIRST_DRAWING_max",
153
+ "PREV_DAYS_FIRST_DRAWING_sum": "PREV_DAYS_FIRST_DRAWING_sum",
154
+ "PREV_DAYS_FIRST_DUE_min": "PREV_DAYS_FIRST_DUE_min",
155
+ "PREV_DAYS_LAST_DUE_mean": "PREV_DAYS_LAST_DUE_mean",
156
+ "PREV_DAYS_LAST_DUE_min": "PREV_DAYS_LAST_DUE_min",
157
+ "PREV_DAYS_TERMINATION_mean": "PREV_DAYS_TERMINATION_mean",
158
+ "PREV_NFLAG_INSURED_ON_APPROVAL_sum": "PREV_NFLAG_INSURED_ON_APPROVAL_sum",
159
+ "PREV_NAME_PAYMENT_TYPE_NUNIQUE": "PREV_NAME_PAYMENT_TYPE_NUNIQUE",
160
+ "PREV_CODE_REJECT_REASON_NUNIQUE": "PREV_CODE_REJECT_REASON_NUNIQUE",
161
+ "PREV_NAME_TYPE_SUITE_NUNIQUE": "PREV_NAME_TYPE_SUITE_NUNIQUE",
162
+ "PREV_NAME_CLIENT_TYPE_NUNIQUE": "PREV_NAME_CLIENT_TYPE_NUNIQUE",
163
+ "PREV_NAME_PRODUCT_TYPE_NUNIQUE": "PREV_NAME_PRODUCT_TYPE_NUNIQUE",
164
+ "PREV_NAME_SELLER_INDUSTRY_NUNIQUE": "PREV_NAME_SELLER_INDUSTRY_NUNIQUE",
165
+ "PREV_NAME_CONTRACT_TYPE_XNA": "PREV_NAME_CONTRACT_TYPE_XNA",
166
+ "PREV_WEEKDAY_APPR_PROCESS_START_FRIDAY": "PREV_WEEKDAY_APPR_PROCESS_START_FRIDAY",
167
+ "PREV_WEEKDAY_APPR_PROCESS_START_MONDAY": "PREV_WEEKDAY_APPR_PROCESS_START_MONDAY",
168
+ "PREV_WEEKDAY_APPR_PROCESS_START_SATURDAY": "PREV_WEEKDAY_APPR_PROCESS_START_SATURDAY",
169
+ "PREV_WEEKDAY_APPR_PROCESS_START_SUNDAY": "PREV_WEEKDAY_APPR_PROCESS_START_SUNDAY",
170
+ "PREV_WEEKDAY_APPR_PROCESS_START_THURSDAY": "PREV_WEEKDAY_APPR_PROCESS_START_THURSDAY",
171
+ "PREV_WEEKDAY_APPR_PROCESS_START_TUESDAY": "PREV_WEEKDAY_APPR_PROCESS_START_TUESDAY",
172
+ "PREV_WEEKDAY_APPR_PROCESS_START_WEDNESDAY": "PREV_WEEKDAY_APPR_PROCESS_START_WEDNESDAY",
173
+ "PREV_FLAG_LAST_APPL_PER_CONTRACT_N": "PREV_FLAG_LAST_APPL_PER_CONTRACT_N",
174
+ "PREV_NAME_CASH_LOAN_PURPOSE_Building_a_house_or_an_annex": "PREV_NAME_CASH_LOAN_PURPOSE_Building a house or an annex",
175
+ "PREV_NAME_CASH_LOAN_PURPOSE_Business_development": "PREV_NAME_CASH_LOAN_PURPOSE_Business development",
176
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_garage": "PREV_NAME_CASH_LOAN_PURPOSE_Buying a garage",
177
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_holiday_home___land": "PREV_NAME_CASH_LOAN_PURPOSE_Buying a holiday home / land",
178
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_home": "PREV_NAME_CASH_LOAN_PURPOSE_Buying a home",
179
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_new_car": "PREV_NAME_CASH_LOAN_PURPOSE_Buying a new car",
180
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_used_car": "PREV_NAME_CASH_LOAN_PURPOSE_Buying a used car",
181
+ "PREV_NAME_CASH_LOAN_PURPOSE_Car_repairs": "PREV_NAME_CASH_LOAN_PURPOSE_Car repairs",
182
+ "PREV_NAME_CASH_LOAN_PURPOSE_Education": "PREV_NAME_CASH_LOAN_PURPOSE_Education",
183
+ "PREV_NAME_CASH_LOAN_PURPOSE_Everyday_expenses": "PREV_NAME_CASH_LOAN_PURPOSE_Everyday expenses",
184
+ "PREV_NAME_CASH_LOAN_PURPOSE_Furniture": "PREV_NAME_CASH_LOAN_PURPOSE_Furniture",
185
+ "PREV_NAME_CASH_LOAN_PURPOSE_Gasification___water_supply": "PREV_NAME_CASH_LOAN_PURPOSE_Gasification / water supply",
186
+ "PREV_NAME_CASH_LOAN_PURPOSE_Hobby": "PREV_NAME_CASH_LOAN_PURPOSE_Hobby",
187
+ "PREV_NAME_CASH_LOAN_PURPOSE_Journey": "PREV_NAME_CASH_LOAN_PURPOSE_Journey",
188
+ "PREV_NAME_CASH_LOAN_PURPOSE_Medicine": "PREV_NAME_CASH_LOAN_PURPOSE_Medicine",
189
+ "PREV_NAME_CASH_LOAN_PURPOSE_Money_for_a_third_person": "PREV_NAME_CASH_LOAN_PURPOSE_Money for a third person",
190
+ "PREV_NAME_CASH_LOAN_PURPOSE_Other": "PREV_NAME_CASH_LOAN_PURPOSE_Other",
191
+ "PREV_NAME_CASH_LOAN_PURPOSE_Payments_on_other_loans": "PREV_NAME_CASH_LOAN_PURPOSE_Payments on other loans",
192
+ "PREV_NAME_CASH_LOAN_PURPOSE_Purchase_of_electronic_equipment": "PREV_NAME_CASH_LOAN_PURPOSE_Purchase of electronic equipment",
193
+ "PREV_NAME_CASH_LOAN_PURPOSE_Refusal_to_name_the_goal": "PREV_NAME_CASH_LOAN_PURPOSE_Refusal to name the goal",
194
+ "PREV_NAME_CASH_LOAN_PURPOSE_Repairs": "PREV_NAME_CASH_LOAN_PURPOSE_Repairs",
195
+ "PREV_NAME_CASH_LOAN_PURPOSE_Urgent_needs": "PREV_NAME_CASH_LOAN_PURPOSE_Urgent needs",
196
+ "PREV_NAME_CASH_LOAN_PURPOSE_Wedding___gift___holiday": "PREV_NAME_CASH_LOAN_PURPOSE_Wedding / gift / holiday",
197
+ "PREV_NAME_CONTRACT_STATUS_Refused": "PREV_NAME_CONTRACT_STATUS_Refused",
198
+ "PREV_NAME_CONTRACT_STATUS_Unused_offer": "PREV_NAME_CONTRACT_STATUS_Unused offer",
199
+ "PREV_NAME_PAYMENT_TYPE_Cashless_from_the_account_of_the_employer": "PREV_NAME_PAYMENT_TYPE_Cashless from the account of the employer",
200
+ "PREV_NAME_PAYMENT_TYPE_Non_cash_from_your_account": "PREV_NAME_PAYMENT_TYPE_Non-cash from your account",
201
+ "PREV_CODE_REJECT_REASON_LIMIT": "PREV_CODE_REJECT_REASON_LIMIT",
202
+ "PREV_CODE_REJECT_REASON_SCO": "PREV_CODE_REJECT_REASON_SCO",
203
+ "PREV_CODE_REJECT_REASON_SCOFR": "PREV_CODE_REJECT_REASON_SCOFR",
204
+ "PREV_CODE_REJECT_REASON_SYSTEM": "PREV_CODE_REJECT_REASON_SYSTEM",
205
+ "PREV_CODE_REJECT_REASON_VERIF": "PREV_CODE_REJECT_REASON_VERIF",
206
+ "PREV_CODE_REJECT_REASON_XNA": "PREV_CODE_REJECT_REASON_XNA",
207
+ "PREV_NAME_TYPE_SUITE_Children": "PREV_NAME_TYPE_SUITE_Children",
208
+ "PREV_NAME_TYPE_SUITE_Family": "PREV_NAME_TYPE_SUITE_Family",
209
+ "PREV_NAME_TYPE_SUITE_Group_of_people": "PREV_NAME_TYPE_SUITE_Group of people",
210
+ "PREV_NAME_TYPE_SUITE_Other_A": "PREV_NAME_TYPE_SUITE_Other_A",
211
+ "PREV_NAME_TYPE_SUITE_Other_B": "PREV_NAME_TYPE_SUITE_Other_B",
212
+ "PREV_NAME_TYPE_SUITE_Spouse__partner": "PREV_NAME_TYPE_SUITE_Spouse, partner",
213
+ "PREV_NAME_CLIENT_TYPE_New": "PREV_NAME_CLIENT_TYPE_New",
214
+ "PREV_NAME_CLIENT_TYPE_Refreshed": "PREV_NAME_CLIENT_TYPE_Refreshed",
215
+ "PREV_NAME_CLIENT_TYPE_XNA": "PREV_NAME_CLIENT_TYPE_XNA",
216
+ "PREV_NAME_GOODS_CATEGORY_Additional_Service": "PREV_NAME_GOODS_CATEGORY_Additional Service",
217
+ "PREV_NAME_GOODS_CATEGORY_Animals": "PREV_NAME_GOODS_CATEGORY_Animals",
218
+ "PREV_NAME_GOODS_CATEGORY_Audio_Video": "PREV_NAME_GOODS_CATEGORY_Audio/Video",
219
+ "PREV_NAME_GOODS_CATEGORY_Auto_Accessories": "PREV_NAME_GOODS_CATEGORY_Auto Accessories",
220
+ "PREV_NAME_GOODS_CATEGORY_Clothing_and_Accessories": "PREV_NAME_GOODS_CATEGORY_Clothing and Accessories",
221
+ "PREV_NAME_GOODS_CATEGORY_Computers": "PREV_NAME_GOODS_CATEGORY_Computers",
222
+ "PREV_NAME_GOODS_CATEGORY_Consumer_Electronics": "PREV_NAME_GOODS_CATEGORY_Consumer Electronics",
223
+ "PREV_NAME_GOODS_CATEGORY_Direct_Sales": "PREV_NAME_GOODS_CATEGORY_Direct Sales",
224
+ "PREV_NAME_GOODS_CATEGORY_Education": "PREV_NAME_GOODS_CATEGORY_Education",
225
+ "PREV_NAME_GOODS_CATEGORY_Fitness": "PREV_NAME_GOODS_CATEGORY_Fitness",
226
+ "PREV_NAME_GOODS_CATEGORY_Gardening": "PREV_NAME_GOODS_CATEGORY_Gardening",
227
+ "PREV_NAME_GOODS_CATEGORY_Homewares": "PREV_NAME_GOODS_CATEGORY_Homewares",
228
+ "PREV_NAME_GOODS_CATEGORY_Insurance": "PREV_NAME_GOODS_CATEGORY_Insurance",
229
+ "PREV_NAME_GOODS_CATEGORY_Jewelry": "PREV_NAME_GOODS_CATEGORY_Jewelry",
230
+ "PREV_NAME_GOODS_CATEGORY_Medical_Supplies": "PREV_NAME_GOODS_CATEGORY_Medical Supplies",
231
+ "PREV_NAME_GOODS_CATEGORY_Medicine": "PREV_NAME_GOODS_CATEGORY_Medicine",
232
+ "PREV_NAME_GOODS_CATEGORY_Office_Appliances": "PREV_NAME_GOODS_CATEGORY_Office Appliances",
233
+ "PREV_NAME_GOODS_CATEGORY_Other": "PREV_NAME_GOODS_CATEGORY_Other",
234
+ "PREV_NAME_GOODS_CATEGORY_Photo___Cinema_Equipment": "PREV_NAME_GOODS_CATEGORY_Photo / Cinema Equipment",
235
+ "PREV_NAME_GOODS_CATEGORY_Sport_and_Leisure": "PREV_NAME_GOODS_CATEGORY_Sport and Leisure",
236
+ "PREV_NAME_GOODS_CATEGORY_Tourism": "PREV_NAME_GOODS_CATEGORY_Tourism",
237
+ "PREV_NAME_GOODS_CATEGORY_Vehicles": "PREV_NAME_GOODS_CATEGORY_Vehicles",
238
+ "PREV_NAME_GOODS_CATEGORY_Weapon": "PREV_NAME_GOODS_CATEGORY_Weapon",
239
+ "PREV_NAME_GOODS_CATEGORY_XNA": "PREV_NAME_GOODS_CATEGORY_XNA",
240
+ "PREV_NAME_PORTFOLIO_POS": "PREV_NAME_PORTFOLIO_POS",
241
+ "PREV_NAME_PRODUCT_TYPE_walk_in": "PREV_NAME_PRODUCT_TYPE_walk-in",
242
+ "PREV_CHANNEL_TYPE_AP___Cash_loan_": "PREV_CHANNEL_TYPE_AP+ (Cash loan)",
243
+ "PREV_CHANNEL_TYPE_Car_dealer": "PREV_CHANNEL_TYPE_Car dealer",
244
+ "PREV_CHANNEL_TYPE_Channel_of_corporate_sales": "PREV_CHANNEL_TYPE_Channel of corporate sales",
245
+ "PREV_CHANNEL_TYPE_Contact_center": "PREV_CHANNEL_TYPE_Contact center",
246
+ "PREV_CHANNEL_TYPE_Regional___Local": "PREV_CHANNEL_TYPE_Regional / Local",
247
+ "PREV_CHANNEL_TYPE_Stone": "PREV_CHANNEL_TYPE_Stone",
248
+ "PREV_NAME_SELLER_INDUSTRY_Auto_technology": "PREV_NAME_SELLER_INDUSTRY_Auto technology",
249
+ "PREV_NAME_SELLER_INDUSTRY_Connectivity": "PREV_NAME_SELLER_INDUSTRY_Connectivity",
250
+ "PREV_NAME_SELLER_INDUSTRY_Construction": "PREV_NAME_SELLER_INDUSTRY_Construction",
251
+ "PREV_NAME_SELLER_INDUSTRY_Industry": "PREV_NAME_SELLER_INDUSTRY_Industry",
252
+ "PREV_NAME_SELLER_INDUSTRY_Jewelry": "PREV_NAME_SELLER_INDUSTRY_Jewelry",
253
+ "PREV_NAME_SELLER_INDUSTRY_MLM_partners": "PREV_NAME_SELLER_INDUSTRY_MLM partners",
254
+ "PREV_NAME_SELLER_INDUSTRY_Tourism": "PREV_NAME_SELLER_INDUSTRY_Tourism",
255
+ "PREV_NAME_YIELD_GROUP_XNA": "PREV_NAME_YIELD_GROUP_XNA",
256
+ "PREV_NAME_YIELD_GROUP_low_action": "PREV_NAME_YIELD_GROUP_low_action",
257
+ "PREV_NAME_YIELD_GROUP_low_normal": "PREV_NAME_YIELD_GROUP_low_normal",
258
+ "PREV_PRODUCT_COMBINATION_Card_Street": "PREV_PRODUCT_COMBINATION_Card Street",
259
+ "PREV_PRODUCT_COMBINATION_Cash_Street__high": "PREV_PRODUCT_COMBINATION_Cash Street: high",
260
+ "PREV_PRODUCT_COMBINATION_Cash_Street__low": "PREV_PRODUCT_COMBINATION_Cash Street: low",
261
+ "PREV_PRODUCT_COMBINATION_Cash_Street__middle": "PREV_PRODUCT_COMBINATION_Cash Street: middle",
262
+ "PREV_PRODUCT_COMBINATION_Cash_X_Sell__high": "PREV_PRODUCT_COMBINATION_Cash X-Sell: high",
263
+ "PREV_PRODUCT_COMBINATION_Cash_X_Sell__middle": "PREV_PRODUCT_COMBINATION_Cash X-Sell: middle",
264
+ "PREV_PRODUCT_COMBINATION_POS_industry_with_interest": "PREV_PRODUCT_COMBINATION_POS industry with interest",
265
+ "PREV_PRODUCT_COMBINATION_POS_industry_without_interest": "PREV_PRODUCT_COMBINATION_POS industry without interest",
266
+ "PREV_PRODUCT_COMBINATION_POS_mobile_without_interest": "PREV_PRODUCT_COMBINATION_POS mobile without interest",
267
+ "PREV_PRODUCT_COMBINATION_POS_other_with_interest": "PREV_PRODUCT_COMBINATION_POS other with interest",
268
+ "PREV_PRODUCT_COMBINATION_POS_others_without_interest": "PREV_PRODUCT_COMBINATION_POS others without interest",
269
+ "INST_PAY_NUM_INSTALMENT_VERSION_mean": "INST_PAY_NUM_INSTALMENT_VERSION_mean",
270
+ "INST_PAY_DAYS_ENTRY_PAYMENT_mean": "INST_PAY_DAYS_ENTRY_PAYMENT_mean",
271
+ "INST_PAY_DAYS_ENTRY_PAYMENT_min": "INST_PAY_DAYS_ENTRY_PAYMENT_min",
272
+ "INST_PAY_AMT_PAYMENT_mean": "INST_PAY_AMT_PAYMENT_mean",
273
+ "INST_PAY_AMT_PAYMENT_min": "INST_PAY_AMT_PAYMENT_min",
274
+ "INST_PAY_AMT_PAYMENT_sum": "INST_PAY_AMT_PAYMENT_sum",
275
+ "INST_PAY_PAYMENT_DELAY_mean": "INST_PAY_PAYMENT_DELAY_mean",
276
+ "INST_PAY_PAYMENT_DELAY_max": "INST_PAY_PAYMENT_DELAY_max",
277
+ "INST_PAY_PAYMENT_DELAY_min": "INST_PAY_PAYMENT_DELAY_min",
278
+ "INST_PAY_PAYMENT_DELAY_sum": "INST_PAY_PAYMENT_DELAY_sum",
279
+ "INST_PAY_PAYMENT_RATIO_mean": "INST_PAY_PAYMENT_RATIO_mean",
280
+ "INST_PAY_PAYMENT_RATIO_min": "INST_PAY_PAYMENT_RATIO_min",
281
+ "CC_MONTHS_BALANCE_mean": "CC_MONTHS_BALANCE_mean",
282
+ "CC_MONTHS_BALANCE_max": "CC_MONTHS_BALANCE_max",
283
+ "CC_AMT_BALANCE_mean": "CC_AMT_BALANCE_mean",
284
+ "CC_AMT_CREDIT_LIMIT_ACTUAL_max": "CC_AMT_CREDIT_LIMIT_ACTUAL_max",
285
+ "CC_AMT_CREDIT_LIMIT_ACTUAL_sum": "CC_AMT_CREDIT_LIMIT_ACTUAL_sum",
286
+ "CC_AMT_DRAWINGS_ATM_CURRENT_mean": "CC_AMT_DRAWINGS_ATM_CURRENT_mean",
287
+ "CC_AMT_DRAWINGS_ATM_CURRENT_sum": "CC_AMT_DRAWINGS_ATM_CURRENT_sum",
288
+ "CC_AMT_DRAWINGS_CURRENT_mean": "CC_AMT_DRAWINGS_CURRENT_mean",
289
+ "CC_AMT_DRAWINGS_CURRENT_max": "CC_AMT_DRAWINGS_CURRENT_max",
290
+ "CC_AMT_DRAWINGS_CURRENT_min": "CC_AMT_DRAWINGS_CURRENT_min",
291
+ "CC_AMT_DRAWINGS_POS_CURRENT_min": "CC_AMT_DRAWINGS_POS_CURRENT_min",
292
+ "CC_AMT_INST_MIN_REGULARITY_min": "CC_AMT_INST_MIN_REGULARITY_min",
293
+ "CC_AMT_PAYMENT_CURRENT_min": "CC_AMT_PAYMENT_CURRENT_min",
294
+ "CC_AMT_PAYMENT_TOTAL_CURRENT_max": "CC_AMT_PAYMENT_TOTAL_CURRENT_max",
295
+ "CC_CNT_DRAWINGS_ATM_CURRENT_mean": "CC_CNT_DRAWINGS_ATM_CURRENT_mean",
296
+ "CC_CNT_DRAWINGS_ATM_CURRENT_min": "CC_CNT_DRAWINGS_ATM_CURRENT_min",
297
+ "CC_CNT_DRAWINGS_ATM_CURRENT_sum": "CC_CNT_DRAWINGS_ATM_CURRENT_sum",
298
+ "CC_CNT_DRAWINGS_CURRENT_max": "CC_CNT_DRAWINGS_CURRENT_max",
299
+ "CC_CNT_DRAWINGS_CURRENT_min": "CC_CNT_DRAWINGS_CURRENT_min",
300
+ "CC_CNT_DRAWINGS_OTHER_CURRENT_mean": "CC_CNT_DRAWINGS_OTHER_CURRENT_mean",
301
+ "CC_CNT_DRAWINGS_OTHER_CURRENT_min": "CC_CNT_DRAWINGS_OTHER_CURRENT_min",
302
+ "CC_SK_DPD_max": "CC_SK_DPD_max",
303
+ "CC_SK_DPD_DEF_max": "CC_SK_DPD_DEF_max",
304
+ "CC_UTILIZATION_mean": "CC_UTILIZATION_mean",
305
+ "CC_UTILIZATION_max": "CC_UTILIZATION_max",
306
+ "CC_UTILIZATION_min": "CC_UTILIZATION_min",
307
+ "CC_PAYMENT_RATIO_max": "CC_PAYMENT_RATIO_max",
308
+ "CC_PAYMENT_RATIO_min": "CC_PAYMENT_RATIO_min",
309
+ "CC_NAME_CONTRACT_STATUS_NUNIQUE": "CC_NAME_CONTRACT_STATUS_NUNIQUE",
310
+ "CC_NAME_CONTRACT_STATUS_Approved": "CC_NAME_CONTRACT_STATUS_Approved",
311
+ "CC_NAME_CONTRACT_STATUS_Completed": "CC_NAME_CONTRACT_STATUS_Completed",
312
+ "CC_NAME_CONTRACT_STATUS_Refused": "CC_NAME_CONTRACT_STATUS_Refused",
313
+ "CC_NAME_CONTRACT_STATUS_Sent_proposal": "CC_NAME_CONTRACT_STATUS_Sent proposal",
314
+ "CC_NAME_CONTRACT_STATUS_Signed": "CC_NAME_CONTRACT_STATUS_Signed",
315
+ "POS_CNT_INSTALMENT_min": "POS_CNT_INSTALMENT_min",
316
+ "POS_CNT_INSTALMENT_FUTURE_min": "POS_CNT_INSTALMENT_FUTURE_min",
317
+ "POS_SK_DPD_mean": "POS_SK_DPD_mean",
318
+ "POS_SK_DPD_min": "POS_SK_DPD_min",
319
+ "POS_SK_DPD_DEF_max": "POS_SK_DPD_DEF_max",
320
+ "POS_SK_DPD_DEF_min": "POS_SK_DPD_DEF_min",
321
+ "POS_NAME_CONTRACT_STATUS_NUNIQUE": "POS_NAME_CONTRACT_STATUS_NUNIQUE",
322
+ "POS_NAME_CONTRACT_STATUS_Amortized_debt": "POS_NAME_CONTRACT_STATUS_Amortized debt",
323
+ "POS_NAME_CONTRACT_STATUS_Approved": "POS_NAME_CONTRACT_STATUS_Approved",
324
+ "POS_NAME_CONTRACT_STATUS_Canceled": "POS_NAME_CONTRACT_STATUS_Canceled",
325
+ "POS_NAME_CONTRACT_STATUS_Completed": "POS_NAME_CONTRACT_STATUS_Completed",
326
+ "POS_NAME_CONTRACT_STATUS_Returned_to_the_store": "POS_NAME_CONTRACT_STATUS_Returned to the store",
327
+ "POS_NAME_CONTRACT_STATUS_Signed": "POS_NAME_CONTRACT_STATUS_Signed",
328
+ "POS_NAME_CONTRACT_STATUS_XNA": "POS_NAME_CONTRACT_STATUS_XNA"
329
+ }
App/generate_example.py ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Générer un example de donnée pour un client
2
+ import pandas as pd
3
+ import json
4
+ import joblib
5
+ import numpy as np
6
+ import re
7
+ from typing import Optional
8
+
9
+ df = joblib.load("./data/app_test_clean_v2.joblib")
10
+
11
+ def sanitize(name: str) -> str:
12
+ """
13
+ Transforme un nom de colonne en identifiant Python valide.
14
+ """
15
+ # Remplacer tout caractère non alphanumérique par un underscore
16
+ name = re.sub(r'[^0-9a-zA-Z_]', '_', name)
17
+
18
+ # Si le nom commence par un chiffre → préfixer
19
+ if re.match(r'^[0-9]', name):
20
+ name = f"col_{name}"
21
+
22
+ return name
23
+
24
+ df.columns = [sanitize(c) for c in df.columns]
25
+
26
+ # Prendre un individu au hasard
27
+ example = df.sample(1).iloc[0].to_dict()
28
+ for k, v in example.items():
29
+ if isinstance(v, float) and (np.isnan(v)):
30
+ example[k] = None
31
+
32
+ fields = {}
33
+
34
+ # Générer automatiquement les types
35
+ for col, dtype in df.dtypes.items():
36
+ if dtype == bool or df[col].dropna().isin([0, 1, True, False]).all():
37
+ fields[col] = (Optional[bool], None)
38
+
39
+ elif "int" in str(dtype):
40
+ fields[col] = (Optional[int], None)
41
+
42
+ elif "float" in str(dtype):
43
+ fields[col] = (Optional[float], None)
44
+
45
+ else:
46
+ fields[col] = (Optional[str], None)
47
+
48
+ # Afficher sous format JSON
49
+ print(json.dumps(example, indent=4))
App/generate_pydantic_model.py ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import joblib
2
+ import numpy as np
3
+ import re
4
+
5
+ df = joblib.load("./data/app_test_clean_v2.joblib")
6
+
7
+ def sanitize(name: str) -> str:
8
+ """
9
+ Transforme un nom de colonne en identifiant Python valide.
10
+ """
11
+ # Remplacer tout caractère non alphanumérique par un underscore
12
+ name = re.sub(r'[^0-9a-zA-Z_]', '_', name)
13
+
14
+ # Si le nom commence par un chiffre → préfixer
15
+ if re.match(r'^[0-9]', name):
16
+ name = f"col_{name}"
17
+
18
+ return name
19
+
20
+ df.columns = [sanitize(c) for c in df.columns]
21
+
22
+
23
+ example = df.sample(1).iloc[0].to_dict()
24
+ for k, v in example.items():
25
+ if isinstance(v, float) and (np.isnan(v)):
26
+ example[k] = None
27
+
28
+ fields = []
29
+
30
+ for col, dtype in df.dtypes.items():
31
+ clean_col = sanitize(col)
32
+
33
+ if dtype == bool or df[col].dropna().isin([0, 1, True, False]).all():
34
+ py_type = "Optional[bool]"
35
+ elif "int" in str(dtype):
36
+ py_type = "Optional[int]"
37
+ elif "float" in str(dtype):
38
+ py_type = "Optional[float]"
39
+ else:
40
+ py_type = "Optional[str]"
41
+
42
+ fields.append(f" {clean_col}: {py_type} = None")
43
+
44
+
45
+ model_code = f"""
46
+ from pydantic import BaseModel
47
+ from typing import Optional
48
+
49
+ class ClientFeatures(BaseModel):
50
+ {chr(10).join(fields)}
51
+ """
52
+
53
+ with open("App/models.py", "w") as f:
54
+ f.write(model_code)
55
+
56
+ print("✔️ Modèle Pydantic généré dans App/models.py")
57
+
58
+ mapping = {sanitize(c): c for c in joblib.load("./data/app_test_clean_v2.joblib").columns}
59
+
60
+ import json
61
+ with open("App/column_mapping.json", "w") as f:
62
+ json.dump(mapping, f, indent=4)
63
+
App/main.py ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #
2
+ from fastapi import FastAPI, HTTPException
3
+ import joblib
4
+ import pandas as pd
5
+ import json
6
+ from App.models import ClientFeatures
7
+
8
+ app = FastAPI()
9
+
10
+ # Charger le pipeline MLflow (attend les colonnes ORIGINALES)
11
+ pipe = joblib.load("./BestModel/pipeline_complet.joblib")
12
+
13
+ # Charger le seuil optimal
14
+ threshold = joblib.load("./BestModel/best_threshold.joblib")
15
+
16
+ # Charger le mapping clean -> original
17
+ with open("App/column_mapping.json") as f:
18
+ COLUMN_MAPPING = json.load(f)
19
+
20
+
21
+ @app.post("/predict")
22
+ def predict(features: ClientFeatures):
23
+
24
+ # Validation métier
25
+ data = features.dict()
26
+
27
+ # Âge invalide
28
+ if "DAYS_BIRTH" in data and data["DAYS_BIRTH"] is not None:
29
+ # Test attend : âge impossible → 422 + "Âge invalide"
30
+ if data["DAYS_BIRTH"] > 0 or data["DAYS_BIRTH"] > -18*365:
31
+ raise HTTPException(
32
+ status_code=422,
33
+ detail="Âge invalide")
34
+
35
+ # Revenu invalide
36
+ if "AMT_INCOME_TOTAL" in data and data["AMT_INCOME_TOTAL"] is not None:
37
+ if data["AMT_INCOME_TOTAL"] <= 0:
38
+ raise HTTPException(
39
+ status_code=422,
40
+ detail="Revenu invalide")
41
+
42
+ # Conversion en DataFrame
43
+ df = pd.DataFrame([data])
44
+
45
+ # Remapping clean à l'original
46
+ df = df.rename(columns=COLUMN_MAPPING)
47
+
48
+ # Vérifier que toutes les colonnes attendues par MLflow sont présentes
49
+ missing = set(pipe.feature_names_in_) - set(df.columns)
50
+ if missing:
51
+ raise HTTPException(
52
+ status_code=422,
53
+ detail=f"Colonnes manquantes après remapping : {missing}")
54
+
55
+ # Prédiction MLflow
56
+ score = pipe.predict_proba(df)[0][1]
57
+ decision = "ACCORDÉ" if score < threshold else "REFUSÉ"
58
+
59
+ return {
60
+ "score": float(score),
61
+ "decision": decision,
62
+ "threshold": float(threshold)
63
+ }
App/models.py ADDED
@@ -0,0 +1,667 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ from pydantic import BaseModel
3
+ from typing import Optional
4
+
5
+ class ClientFeatures(BaseModel):
6
+ SK_ID_CURR: Optional[int] = None
7
+ NAME_CONTRACT_TYPE: Optional[str] = None
8
+ CODE_GENDER: Optional[str] = None
9
+ FLAG_OWN_CAR: Optional[str] = None
10
+ FLAG_OWN_REALTY: Optional[str] = None
11
+ CNT_CHILDREN: Optional[int] = None
12
+ AMT_INCOME_TOTAL: Optional[float] = None
13
+ AMT_GOODS_PRICE: Optional[float] = None
14
+ NAME_TYPE_SUITE: Optional[str] = None
15
+ NAME_INCOME_TYPE: Optional[str] = None
16
+ NAME_EDUCATION_TYPE: Optional[str] = None
17
+ NAME_FAMILY_STATUS: Optional[str] = None
18
+ NAME_HOUSING_TYPE: Optional[str] = None
19
+ REGION_POPULATION_RELATIVE: Optional[float] = None
20
+ DAYS_BIRTH: Optional[int] = None
21
+ DAYS_EMPLOYED: Optional[int] = None
22
+ DAYS_REGISTRATION: Optional[float] = None
23
+ DAYS_ID_PUBLISH: Optional[int] = None
24
+ OWN_CAR_AGE: Optional[float] = None
25
+ FLAG_MOBIL: Optional[bool] = None
26
+ FLAG_EMP_PHONE: Optional[bool] = None
27
+ FLAG_WORK_PHONE: Optional[bool] = None
28
+ FLAG_CONT_MOBILE: Optional[bool] = None
29
+ FLAG_PHONE: Optional[bool] = None
30
+ FLAG_EMAIL: Optional[bool] = None
31
+ OCCUPATION_TYPE: Optional[str] = None
32
+ REGION_RATING_CLIENT_W_CITY: Optional[int] = None
33
+ WEEKDAY_APPR_PROCESS_START: Optional[str] = None
34
+ HOUR_APPR_PROCESS_START: Optional[int] = None
35
+ REG_REGION_NOT_LIVE_REGION: Optional[bool] = None
36
+ REG_REGION_NOT_WORK_REGION: Optional[bool] = None
37
+ LIVE_REGION_NOT_WORK_REGION: Optional[bool] = None
38
+ REG_CITY_NOT_LIVE_CITY: Optional[bool] = None
39
+ REG_CITY_NOT_WORK_CITY: Optional[bool] = None
40
+ LIVE_CITY_NOT_WORK_CITY: Optional[bool] = None
41
+ ORGANIZATION_TYPE: Optional[str] = None
42
+ EXT_SOURCE_1: Optional[float] = None
43
+ EXT_SOURCE_2: Optional[float] = None
44
+ EXT_SOURCE_3: Optional[float] = None
45
+ BASEMENTAREA_AVG: Optional[float] = None
46
+ ELEVATORS_AVG: Optional[float] = None
47
+ ENTRANCES_AVG: Optional[float] = None
48
+ FLOORSMAX_AVG: Optional[float] = None
49
+ NONLIVINGAPARTMENTS_AVG: Optional[float] = None
50
+ NONLIVINGAREA_AVG: Optional[float] = None
51
+ YEARS_BEGINEXPLUATATION_MEDI: Optional[float] = None
52
+ YEARS_BUILD_MEDI: Optional[float] = None
53
+ COMMONAREA_MEDI: Optional[float] = None
54
+ LANDAREA_MEDI: Optional[float] = None
55
+ FONDKAPREMONT_MODE: Optional[str] = None
56
+ HOUSETYPE_MODE: Optional[str] = None
57
+ WALLSMATERIAL_MODE: Optional[str] = None
58
+ EMERGENCYSTATE_MODE: Optional[str] = None
59
+ OBS_30_CNT_SOCIAL_CIRCLE: Optional[float] = None
60
+ DEF_30_CNT_SOCIAL_CIRCLE: Optional[float] = None
61
+ DAYS_LAST_PHONE_CHANGE: Optional[float] = None
62
+ FLAG_DOCUMENT_2: Optional[bool] = None
63
+ FLAG_DOCUMENT_3: Optional[bool] = None
64
+ FLAG_DOCUMENT_4: Optional[bool] = None
65
+ FLAG_DOCUMENT_5: Optional[bool] = None
66
+ FLAG_DOCUMENT_6: Optional[bool] = None
67
+ FLAG_DOCUMENT_7: Optional[bool] = None
68
+ FLAG_DOCUMENT_8: Optional[bool] = None
69
+ FLAG_DOCUMENT_9: Optional[bool] = None
70
+ FLAG_DOCUMENT_10: Optional[bool] = None
71
+ FLAG_DOCUMENT_11: Optional[bool] = None
72
+ FLAG_DOCUMENT_12: Optional[bool] = None
73
+ FLAG_DOCUMENT_13: Optional[bool] = None
74
+ FLAG_DOCUMENT_14: Optional[bool] = None
75
+ FLAG_DOCUMENT_15: Optional[bool] = None
76
+ FLAG_DOCUMENT_16: Optional[bool] = None
77
+ FLAG_DOCUMENT_17: Optional[bool] = None
78
+ FLAG_DOCUMENT_18: Optional[bool] = None
79
+ FLAG_DOCUMENT_19: Optional[bool] = None
80
+ FLAG_DOCUMENT_20: Optional[bool] = None
81
+ FLAG_DOCUMENT_21: Optional[bool] = None
82
+ AMT_REQ_CREDIT_BUREAU_HOUR: Optional[float] = None
83
+ AMT_REQ_CREDIT_BUREAU_DAY: Optional[float] = None
84
+ AMT_REQ_CREDIT_BUREAU_WEEK: Optional[float] = None
85
+ AMT_REQ_CREDIT_BUREAU_MON: Optional[float] = None
86
+ AMT_REQ_CREDIT_BUREAU_QRT: Optional[float] = None
87
+ BUREAU_SK_ID_BUREAU_max: Optional[float] = None
88
+ BUREAU_DAYS_CREDIT_mean: Optional[float] = None
89
+ BUREAU_DAYS_CREDIT_max: Optional[float] = None
90
+ BUREAU_DAYS_CREDIT_min: Optional[float] = None
91
+ BUREAU_CREDIT_DAY_OVERDUE_mean: Optional[float] = None
92
+ BUREAU_CREDIT_DAY_OVERDUE_sum: Optional[float] = None
93
+ BUREAU_DAYS_CREDIT_ENDDATE_mean: Optional[float] = None
94
+ BUREAU_DAYS_CREDIT_ENDDATE_max: Optional[float] = None
95
+ BUREAU_DAYS_CREDIT_ENDDATE_min: Optional[float] = None
96
+ BUREAU_DAYS_CREDIT_ENDDATE_sum: Optional[float] = None
97
+ BUREAU_DAYS_ENDDATE_FACT_max: Optional[float] = None
98
+ BUREAU_DAYS_ENDDATE_FACT_sum: Optional[float] = None
99
+ BUREAU_AMT_CREDIT_MAX_OVERDUE_max: Optional[float] = None
100
+ BUREAU_CNT_CREDIT_PROLONG_max: Optional[float] = None
101
+ BUREAU_CNT_CREDIT_PROLONG_sum: Optional[float] = None
102
+ BUREAU_AMT_CREDIT_SUM_mean: Optional[float] = None
103
+ BUREAU_AMT_CREDIT_SUM_min: Optional[float] = None
104
+ BUREAU_AMT_CREDIT_SUM_sum: Optional[float] = None
105
+ BUREAU_AMT_CREDIT_SUM_DEBT_max: Optional[float] = None
106
+ BUREAU_AMT_CREDIT_SUM_DEBT_sum: Optional[float] = None
107
+ BUREAU_AMT_CREDIT_SUM_LIMIT_mean: Optional[float] = None
108
+ BUREAU_AMT_CREDIT_SUM_LIMIT_sum: Optional[float] = None
109
+ BUREAU_AMT_CREDIT_SUM_OVERDUE_max: Optional[float] = None
110
+ BUREAU_AMT_CREDIT_SUM_OVERDUE_sum: Optional[float] = None
111
+ BUREAU_AMT_ANNUITY_mean: Optional[float] = None
112
+ BUREAU_AMT_ANNUITY_min: Optional[float] = None
113
+ BUREAU_MONTHS_BALANCE_max: Optional[float] = None
114
+ BUREAU_CREDIT_ACTIVE_NUNIQUE: Optional[float] = None
115
+ BUREAU_CREDIT_CURRENCY_NUNIQUE: Optional[float] = None
116
+ BUREAU_CREDIT_TYPE_NUNIQUE: Optional[float] = None
117
+ BUREAU_STATUS_NUNIQUE: Optional[float] = None
118
+ BUREAU_CREDIT_ACTIVE_Bad_debt: Optional[float] = None
119
+ BUREAU_CREDIT_ACTIVE_Sold: Optional[float] = None
120
+ BUREAU_CREDIT_CURRENCY_currency_2: Optional[float] = None
121
+ BUREAU_CREDIT_CURRENCY_currency_3: Optional[float] = None
122
+ BUREAU_CREDIT_CURRENCY_currency_4: Optional[float] = None
123
+ BUREAU_CREDIT_TYPE_Another_type_of_loan: Optional[float] = None
124
+ BUREAU_CREDIT_TYPE_Car_loan: Optional[float] = None
125
+ BUREAU_CREDIT_TYPE_Cash_loan__non_earmarked_: Optional[float] = None
126
+ BUREAU_CREDIT_TYPE_Interbank_credit: Optional[bool] = None
127
+ BUREAU_CREDIT_TYPE_Loan_for_business_development: Optional[float] = None
128
+ BUREAU_CREDIT_TYPE_Loan_for_purchase_of_shares__margin_lending_: Optional[bool] = None
129
+ BUREAU_CREDIT_TYPE_Loan_for_the_purchase_of_equipment: Optional[float] = None
130
+ BUREAU_CREDIT_TYPE_Loan_for_working_capital_replenishment: Optional[float] = None
131
+ BUREAU_CREDIT_TYPE_Microloan: Optional[float] = None
132
+ BUREAU_CREDIT_TYPE_Mobile_operator_loan: Optional[bool] = None
133
+ BUREAU_CREDIT_TYPE_Mortgage: Optional[float] = None
134
+ BUREAU_CREDIT_TYPE_Real_estate_loan: Optional[float] = None
135
+ BUREAU_CREDIT_TYPE_Unknown_type_of_loan: Optional[float] = None
136
+ BUREAU_STATUS_1: Optional[float] = None
137
+ BUREAU_STATUS_2: Optional[float] = None
138
+ BUREAU_STATUS_4: Optional[float] = None
139
+ BUREAU_STATUS_5: Optional[float] = None
140
+ PREV_AMT_ANNUITY_mean: Optional[float] = None
141
+ PREV_AMT_ANNUITY_min: Optional[float] = None
142
+ PREV_AMT_APPLICATION_min: Optional[float] = None
143
+ PREV_AMT_DOWN_PAYMENT_sum: Optional[float] = None
144
+ PREV_HOUR_APPR_PROCESS_START_mean: Optional[float] = None
145
+ PREV_NFLAG_LAST_APPL_IN_DAY_max: Optional[bool] = None
146
+ PREV_NFLAG_LAST_APPL_IN_DAY_min: Optional[bool] = None
147
+ PREV_RATE_DOWN_PAYMENT_max: Optional[float] = None
148
+ PREV_RATE_DOWN_PAYMENT_sum: Optional[float] = None
149
+ PREV_RATE_INTEREST_PRIMARY_max: Optional[float] = None
150
+ PREV_RATE_INTEREST_PRIMARY_sum: Optional[float] = None
151
+ PREV_RATE_INTEREST_PRIVILEGED_max: Optional[float] = None
152
+ PREV_SELLERPLACE_AREA_sum: Optional[float] = None
153
+ PREV_CNT_PAYMENT_max: Optional[float] = None
154
+ PREV_CNT_PAYMENT_min: Optional[float] = None
155
+ PREV_DAYS_FIRST_DRAWING_mean: Optional[float] = None
156
+ PREV_DAYS_FIRST_DRAWING_max: Optional[float] = None
157
+ PREV_DAYS_FIRST_DRAWING_sum: Optional[float] = None
158
+ PREV_DAYS_FIRST_DUE_min: Optional[float] = None
159
+ PREV_DAYS_LAST_DUE_mean: Optional[float] = None
160
+ PREV_DAYS_LAST_DUE_min: Optional[float] = None
161
+ PREV_DAYS_TERMINATION_mean: Optional[float] = None
162
+ PREV_NFLAG_INSURED_ON_APPROVAL_sum: Optional[float] = None
163
+ PREV_NAME_PAYMENT_TYPE_NUNIQUE: Optional[float] = None
164
+ PREV_CODE_REJECT_REASON_NUNIQUE: Optional[float] = None
165
+ PREV_NAME_TYPE_SUITE_NUNIQUE: Optional[float] = None
166
+ PREV_NAME_CLIENT_TYPE_NUNIQUE: Optional[float] = None
167
+ PREV_NAME_PRODUCT_TYPE_NUNIQUE: Optional[float] = None
168
+ PREV_NAME_SELLER_INDUSTRY_NUNIQUE: Optional[float] = None
169
+ PREV_NAME_CONTRACT_TYPE_XNA: Optional[float] = None
170
+ PREV_WEEKDAY_APPR_PROCESS_START_FRIDAY: Optional[float] = None
171
+ PREV_WEEKDAY_APPR_PROCESS_START_MONDAY: Optional[float] = None
172
+ PREV_WEEKDAY_APPR_PROCESS_START_SATURDAY: Optional[float] = None
173
+ PREV_WEEKDAY_APPR_PROCESS_START_SUNDAY: Optional[float] = None
174
+ PREV_WEEKDAY_APPR_PROCESS_START_THURSDAY: Optional[float] = None
175
+ PREV_WEEKDAY_APPR_PROCESS_START_TUESDAY: Optional[float] = None
176
+ PREV_WEEKDAY_APPR_PROCESS_START_WEDNESDAY: Optional[float] = None
177
+ PREV_FLAG_LAST_APPL_PER_CONTRACT_N: Optional[float] = None
178
+ PREV_NAME_CASH_LOAN_PURPOSE_Building_a_house_or_an_annex: Optional[float] = None
179
+ PREV_NAME_CASH_LOAN_PURPOSE_Business_development: Optional[float] = None
180
+ PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_garage: Optional[float] = None
181
+ PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_holiday_home___land: Optional[float] = None
182
+ PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_home: Optional[float] = None
183
+ PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_new_car: Optional[float] = None
184
+ PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_used_car: Optional[float] = None
185
+ PREV_NAME_CASH_LOAN_PURPOSE_Car_repairs: Optional[float] = None
186
+ PREV_NAME_CASH_LOAN_PURPOSE_Education: Optional[float] = None
187
+ PREV_NAME_CASH_LOAN_PURPOSE_Everyday_expenses: Optional[float] = None
188
+ PREV_NAME_CASH_LOAN_PURPOSE_Furniture: Optional[float] = None
189
+ PREV_NAME_CASH_LOAN_PURPOSE_Gasification___water_supply: Optional[float] = None
190
+ PREV_NAME_CASH_LOAN_PURPOSE_Hobby: Optional[float] = None
191
+ PREV_NAME_CASH_LOAN_PURPOSE_Journey: Optional[float] = None
192
+ PREV_NAME_CASH_LOAN_PURPOSE_Medicine: Optional[float] = None
193
+ PREV_NAME_CASH_LOAN_PURPOSE_Money_for_a_third_person: Optional[bool] = None
194
+ PREV_NAME_CASH_LOAN_PURPOSE_Other: Optional[float] = None
195
+ PREV_NAME_CASH_LOAN_PURPOSE_Payments_on_other_loans: Optional[float] = None
196
+ PREV_NAME_CASH_LOAN_PURPOSE_Purchase_of_electronic_equipment: Optional[float] = None
197
+ PREV_NAME_CASH_LOAN_PURPOSE_Refusal_to_name_the_goal: Optional[bool] = None
198
+ PREV_NAME_CASH_LOAN_PURPOSE_Repairs: Optional[float] = None
199
+ PREV_NAME_CASH_LOAN_PURPOSE_Urgent_needs: Optional[float] = None
200
+ PREV_NAME_CASH_LOAN_PURPOSE_Wedding___gift___holiday: Optional[float] = None
201
+ PREV_NAME_CONTRACT_STATUS_Refused: Optional[float] = None
202
+ PREV_NAME_CONTRACT_STATUS_Unused_offer: Optional[float] = None
203
+ PREV_NAME_PAYMENT_TYPE_Cashless_from_the_account_of_the_employer: Optional[float] = None
204
+ PREV_NAME_PAYMENT_TYPE_Non_cash_from_your_account: Optional[float] = None
205
+ PREV_CODE_REJECT_REASON_LIMIT: Optional[float] = None
206
+ PREV_CODE_REJECT_REASON_SCO: Optional[float] = None
207
+ PREV_CODE_REJECT_REASON_SCOFR: Optional[float] = None
208
+ PREV_CODE_REJECT_REASON_SYSTEM: Optional[float] = None
209
+ PREV_CODE_REJECT_REASON_VERIF: Optional[float] = None
210
+ PREV_CODE_REJECT_REASON_XNA: Optional[float] = None
211
+ PREV_NAME_TYPE_SUITE_Children: Optional[float] = None
212
+ PREV_NAME_TYPE_SUITE_Family: Optional[float] = None
213
+ PREV_NAME_TYPE_SUITE_Group_of_people: Optional[float] = None
214
+ PREV_NAME_TYPE_SUITE_Other_A: Optional[float] = None
215
+ PREV_NAME_TYPE_SUITE_Other_B: Optional[float] = None
216
+ PREV_NAME_TYPE_SUITE_Spouse__partner: Optional[float] = None
217
+ PREV_NAME_CLIENT_TYPE_New: Optional[float] = None
218
+ PREV_NAME_CLIENT_TYPE_Refreshed: Optional[float] = None
219
+ PREV_NAME_CLIENT_TYPE_XNA: Optional[float] = None
220
+ PREV_NAME_GOODS_CATEGORY_Additional_Service: Optional[bool] = None
221
+ PREV_NAME_GOODS_CATEGORY_Animals: Optional[bool] = None
222
+ PREV_NAME_GOODS_CATEGORY_Audio_Video: Optional[float] = None
223
+ PREV_NAME_GOODS_CATEGORY_Auto_Accessories: Optional[float] = None
224
+ PREV_NAME_GOODS_CATEGORY_Clothing_and_Accessories: Optional[float] = None
225
+ PREV_NAME_GOODS_CATEGORY_Computers: Optional[float] = None
226
+ PREV_NAME_GOODS_CATEGORY_Consumer_Electronics: Optional[float] = None
227
+ PREV_NAME_GOODS_CATEGORY_Direct_Sales: Optional[float] = None
228
+ PREV_NAME_GOODS_CATEGORY_Education: Optional[float] = None
229
+ PREV_NAME_GOODS_CATEGORY_Fitness: Optional[float] = None
230
+ PREV_NAME_GOODS_CATEGORY_Gardening: Optional[float] = None
231
+ PREV_NAME_GOODS_CATEGORY_Homewares: Optional[float] = None
232
+ PREV_NAME_GOODS_CATEGORY_Insurance: Optional[bool] = None
233
+ PREV_NAME_GOODS_CATEGORY_Jewelry: Optional[float] = None
234
+ PREV_NAME_GOODS_CATEGORY_Medical_Supplies: Optional[float] = None
235
+ PREV_NAME_GOODS_CATEGORY_Medicine: Optional[float] = None
236
+ PREV_NAME_GOODS_CATEGORY_Office_Appliances: Optional[float] = None
237
+ PREV_NAME_GOODS_CATEGORY_Other: Optional[float] = None
238
+ PREV_NAME_GOODS_CATEGORY_Photo___Cinema_Equipment: Optional[float] = None
239
+ PREV_NAME_GOODS_CATEGORY_Sport_and_Leisure: Optional[float] = None
240
+ PREV_NAME_GOODS_CATEGORY_Tourism: Optional[float] = None
241
+ PREV_NAME_GOODS_CATEGORY_Vehicles: Optional[float] = None
242
+ PREV_NAME_GOODS_CATEGORY_Weapon: Optional[float] = None
243
+ PREV_NAME_GOODS_CATEGORY_XNA: Optional[float] = None
244
+ PREV_NAME_PORTFOLIO_POS: Optional[float] = None
245
+ PREV_NAME_PRODUCT_TYPE_walk_in: Optional[float] = None
246
+ PREV_CHANNEL_TYPE_AP___Cash_loan_: Optional[float] = None
247
+ PREV_CHANNEL_TYPE_Car_dealer: Optional[float] = None
248
+ PREV_CHANNEL_TYPE_Channel_of_corporate_sales: Optional[float] = None
249
+ PREV_CHANNEL_TYPE_Contact_center: Optional[float] = None
250
+ PREV_CHANNEL_TYPE_Regional___Local: Optional[float] = None
251
+ PREV_CHANNEL_TYPE_Stone: Optional[float] = None
252
+ PREV_NAME_SELLER_INDUSTRY_Auto_technology: Optional[float] = None
253
+ PREV_NAME_SELLER_INDUSTRY_Connectivity: Optional[float] = None
254
+ PREV_NAME_SELLER_INDUSTRY_Construction: Optional[float] = None
255
+ PREV_NAME_SELLER_INDUSTRY_Industry: Optional[float] = None
256
+ PREV_NAME_SELLER_INDUSTRY_Jewelry: Optional[float] = None
257
+ PREV_NAME_SELLER_INDUSTRY_MLM_partners: Optional[float] = None
258
+ PREV_NAME_SELLER_INDUSTRY_Tourism: Optional[float] = None
259
+ PREV_NAME_YIELD_GROUP_XNA: Optional[float] = None
260
+ PREV_NAME_YIELD_GROUP_low_action: Optional[float] = None
261
+ PREV_NAME_YIELD_GROUP_low_normal: Optional[float] = None
262
+ PREV_PRODUCT_COMBINATION_Card_Street: Optional[float] = None
263
+ PREV_PRODUCT_COMBINATION_Cash_Street__high: Optional[float] = None
264
+ PREV_PRODUCT_COMBINATION_Cash_Street__low: Optional[float] = None
265
+ PREV_PRODUCT_COMBINATION_Cash_Street__middle: Optional[float] = None
266
+ PREV_PRODUCT_COMBINATION_Cash_X_Sell__high: Optional[float] = None
267
+ PREV_PRODUCT_COMBINATION_Cash_X_Sell__middle: Optional[float] = None
268
+ PREV_PRODUCT_COMBINATION_POS_industry_with_interest: Optional[float] = None
269
+ PREV_PRODUCT_COMBINATION_POS_industry_without_interest: Optional[float] = None
270
+ PREV_PRODUCT_COMBINATION_POS_mobile_without_interest: Optional[float] = None
271
+ PREV_PRODUCT_COMBINATION_POS_other_with_interest: Optional[float] = None
272
+ PREV_PRODUCT_COMBINATION_POS_others_without_interest: Optional[float] = None
273
+ INST_PAY_NUM_INSTALMENT_VERSION_mean: Optional[float] = None
274
+ INST_PAY_DAYS_ENTRY_PAYMENT_mean: Optional[float] = None
275
+ INST_PAY_DAYS_ENTRY_PAYMENT_min: Optional[float] = None
276
+ INST_PAY_AMT_PAYMENT_mean: Optional[float] = None
277
+ INST_PAY_AMT_PAYMENT_min: Optional[float] = None
278
+ INST_PAY_AMT_PAYMENT_sum: Optional[float] = None
279
+ INST_PAY_PAYMENT_DELAY_mean: Optional[float] = None
280
+ INST_PAY_PAYMENT_DELAY_max: Optional[float] = None
281
+ INST_PAY_PAYMENT_DELAY_min: Optional[float] = None
282
+ INST_PAY_PAYMENT_DELAY_sum: Optional[float] = None
283
+ INST_PAY_PAYMENT_RATIO_mean: Optional[float] = None
284
+ INST_PAY_PAYMENT_RATIO_min: Optional[float] = None
285
+ CC_MONTHS_BALANCE_mean: Optional[float] = None
286
+ CC_MONTHS_BALANCE_max: Optional[float] = None
287
+ CC_AMT_BALANCE_mean: Optional[float] = None
288
+ CC_AMT_CREDIT_LIMIT_ACTUAL_max: Optional[float] = None
289
+ CC_AMT_CREDIT_LIMIT_ACTUAL_sum: Optional[float] = None
290
+ CC_AMT_DRAWINGS_ATM_CURRENT_mean: Optional[float] = None
291
+ CC_AMT_DRAWINGS_ATM_CURRENT_sum: Optional[float] = None
292
+ CC_AMT_DRAWINGS_CURRENT_mean: Optional[float] = None
293
+ CC_AMT_DRAWINGS_CURRENT_max: Optional[float] = None
294
+ CC_AMT_DRAWINGS_CURRENT_min: Optional[float] = None
295
+ CC_AMT_DRAWINGS_POS_CURRENT_min: Optional[float] = None
296
+ CC_AMT_INST_MIN_REGULARITY_min: Optional[float] = None
297
+ CC_AMT_PAYMENT_CURRENT_min: Optional[float] = None
298
+ CC_AMT_PAYMENT_TOTAL_CURRENT_max: Optional[float] = None
299
+ CC_CNT_DRAWINGS_ATM_CURRENT_mean: Optional[float] = None
300
+ CC_CNT_DRAWINGS_ATM_CURRENT_min: Optional[float] = None
301
+ CC_CNT_DRAWINGS_ATM_CURRENT_sum: Optional[float] = None
302
+ CC_CNT_DRAWINGS_CURRENT_max: Optional[float] = None
303
+ CC_CNT_DRAWINGS_CURRENT_min: Optional[float] = None
304
+ CC_CNT_DRAWINGS_OTHER_CURRENT_mean: Optional[float] = None
305
+ CC_CNT_DRAWINGS_OTHER_CURRENT_min: Optional[bool] = None
306
+ CC_SK_DPD_max: Optional[float] = None
307
+ CC_SK_DPD_DEF_max: Optional[float] = None
308
+ CC_UTILIZATION_mean: Optional[float] = None
309
+ CC_UTILIZATION_max: Optional[float] = None
310
+ CC_UTILIZATION_min: Optional[float] = None
311
+ CC_PAYMENT_RATIO_max: Optional[float] = None
312
+ CC_PAYMENT_RATIO_min: Optional[float] = None
313
+ CC_NAME_CONTRACT_STATUS_NUNIQUE: Optional[float] = None
314
+ CC_NAME_CONTRACT_STATUS_Approved: Optional[bool] = None
315
+ CC_NAME_CONTRACT_STATUS_Completed: Optional[float] = None
316
+ CC_NAME_CONTRACT_STATUS_Refused: Optional[bool] = None
317
+ CC_NAME_CONTRACT_STATUS_Sent_proposal: Optional[bool] = None
318
+ CC_NAME_CONTRACT_STATUS_Signed: Optional[float] = None
319
+ POS_CNT_INSTALMENT_min: Optional[float] = None
320
+ POS_CNT_INSTALMENT_FUTURE_min: Optional[float] = None
321
+ POS_SK_DPD_mean: Optional[float] = None
322
+ POS_SK_DPD_min: Optional[float] = None
323
+ POS_SK_DPD_DEF_max: Optional[float] = None
324
+ POS_SK_DPD_DEF_min: Optional[bool] = None
325
+ POS_NAME_CONTRACT_STATUS_NUNIQUE: Optional[float] = None
326
+ POS_NAME_CONTRACT_STATUS_Amortized_debt: Optional[float] = None
327
+ POS_NAME_CONTRACT_STATUS_Approved: Optional[float] = None
328
+ POS_NAME_CONTRACT_STATUS_Canceled: Optional[bool] = None
329
+ POS_NAME_CONTRACT_STATUS_Completed: Optional[float] = None
330
+ POS_NAME_CONTRACT_STATUS_Returned_to_the_store: Optional[float] = None
331
+ POS_NAME_CONTRACT_STATUS_Signed: Optional[float] = None
332
+ POS_NAME_CONTRACT_STATUS_XNA: Optional[bool] = None
333
+
334
+ # --- Configuration du modèle ---
335
+ model_config = {
336
+ "json_schema_extra": {
337
+ "example": {
338
+ "SK_ID_CURR": 309251,
339
+ "NAME_CONTRACT_TYPE": "Cash loans",
340
+ "CODE_GENDER": "F",
341
+ "FLAG_OWN_CAR": "N",
342
+ "FLAG_OWN_REALTY": "Y",
343
+ "CNT_CHILDREN": 0,
344
+ "AMT_INCOME_TOTAL": 67500.0,
345
+ "AMT_GOODS_PRICE": 225000.0,
346
+ "NAME_TYPE_SUITE": "Unaccompanied",
347
+ "NAME_INCOME_TYPE": "Working",
348
+ "NAME_EDUCATION_TYPE": "Secondary / secondary special",
349
+ "NAME_FAMILY_STATUS": "Married",
350
+ "NAME_HOUSING_TYPE": "House / apartment",
351
+ "REGION_POPULATION_RELATIVE": 0.0145199999999999,
352
+ "DAYS_BIRTH": -12489,
353
+ "DAYS_EMPLOYED": -632,
354
+ "DAYS_REGISTRATION": -5884.0,
355
+ "DAYS_ID_PUBLISH": -127,
356
+ "OWN_CAR_AGE": None,
357
+ "FLAG_MOBIL": True,
358
+ "FLAG_EMP_PHONE": True,
359
+ "FLAG_WORK_PHONE": False,
360
+ "FLAG_CONT_MOBILE": True,
361
+ "FLAG_PHONE": False,
362
+ "FLAG_EMAIL": True,
363
+ "OCCUPATION_TYPE": "Medicine staff",
364
+ "REGION_RATING_CLIENT_W_CITY": 2,
365
+ "WEEKDAY_APPR_PROCESS_START": "TUESDAY",
366
+ "HOUR_APPR_PROCESS_START": 7,
367
+ "REG_REGION_NOT_LIVE_REGION": False,
368
+ "REG_REGION_NOT_WORK_REGION": False,
369
+ "LIVE_REGION_NOT_WORK_REGION": False,
370
+ "REG_CITY_NOT_LIVE_CITY": False,
371
+ "REG_CITY_NOT_WORK_CITY": False,
372
+ "LIVE_CITY_NOT_WORK_CITY": False,
373
+ "ORGANIZATION_TYPE": "Medicine",
374
+ "EXT_SOURCE_1": 0.2991537403878965,
375
+ "EXT_SOURCE_2": 0.3349769787490031,
376
+ "EXT_SOURCE_3": None,
377
+ "BASEMENTAREA_AVG": None,
378
+ "ELEVATORS_AVG": None,
379
+ "ENTRANCES_AVG": None,
380
+ "FLOORSMAX_AVG": None,
381
+ "NONLIVINGAPARTMENTS_AVG": None,
382
+ "NONLIVINGAREA_AVG": None,
383
+ "YEARS_BEGINEXPLUATATION_MEDI": None,
384
+ "YEARS_BUILD_MEDI": None,
385
+ "COMMONAREA_MEDI": None,
386
+ "LANDAREA_MEDI": None,
387
+ "FONDKAPREMONT_MODE": None,
388
+ "HOUSETYPE_MODE": None,
389
+ "WALLSMATERIAL_MODE": None,
390
+ "EMERGENCYSTATE_MODE": None,
391
+ "OBS_30_CNT_SOCIAL_CIRCLE": 0.0,
392
+ "DEF_30_CNT_SOCIAL_CIRCLE": 0.0,
393
+ "DAYS_LAST_PHONE_CHANGE": -581.0,
394
+ "FLAG_DOCUMENT_2": False,
395
+ "FLAG_DOCUMENT_3": True,
396
+ "FLAG_DOCUMENT_4": False,
397
+ "FLAG_DOCUMENT_5": False,
398
+ "FLAG_DOCUMENT_6": False,
399
+ "FLAG_DOCUMENT_7": False,
400
+ "FLAG_DOCUMENT_8": False,
401
+ "FLAG_DOCUMENT_9": False,
402
+ "FLAG_DOCUMENT_10": False,
403
+ "FLAG_DOCUMENT_11": False,
404
+ "FLAG_DOCUMENT_12": False,
405
+ "FLAG_DOCUMENT_13": False,
406
+ "FLAG_DOCUMENT_14": False,
407
+ "FLAG_DOCUMENT_15": False,
408
+ "FLAG_DOCUMENT_16": False,
409
+ "FLAG_DOCUMENT_17": False,
410
+ "FLAG_DOCUMENT_18": False,
411
+ "FLAG_DOCUMENT_19": False,
412
+ "FLAG_DOCUMENT_20": False,
413
+ "FLAG_DOCUMENT_21": False,
414
+ "AMT_REQ_CREDIT_BUREAU_HOUR": 0.0,
415
+ "AMT_REQ_CREDIT_BUREAU_DAY": 0.0,
416
+ "AMT_REQ_CREDIT_BUREAU_WEEK": 0.0,
417
+ "AMT_REQ_CREDIT_BUREAU_MON": 0.0,
418
+ "AMT_REQ_CREDIT_BUREAU_QRT": 0.0,
419
+ "BUREAU_SK_ID_BUREAU_max": 6517759.0,
420
+ "BUREAU_DAYS_CREDIT_mean": -411.0,
421
+ "BUREAU_DAYS_CREDIT_max": -411.0,
422
+ "BUREAU_DAYS_CREDIT_min": -411.0,
423
+ "BUREAU_CREDIT_DAY_OVERDUE_mean": 0.0,
424
+ "BUREAU_CREDIT_DAY_OVERDUE_sum": 0.0,
425
+ "BUREAU_DAYS_CREDIT_ENDDATE_mean": 685.0,
426
+ "BUREAU_DAYS_CREDIT_ENDDATE_max": 685.0,
427
+ "BUREAU_DAYS_CREDIT_ENDDATE_min": 685.0,
428
+ "BUREAU_DAYS_CREDIT_ENDDATE_sum": 9590.0,
429
+ "BUREAU_DAYS_ENDDATE_FACT_max": None,
430
+ "BUREAU_DAYS_ENDDATE_FACT_sum": 0.0,
431
+ "BUREAU_AMT_CREDIT_MAX_OVERDUE_max": 0.0,
432
+ "BUREAU_CNT_CREDIT_PROLONG_max": 0.0,
433
+ "BUREAU_CNT_CREDIT_PROLONG_sum": 0.0,
434
+ "BUREAU_AMT_CREDIT_SUM_mean": 110646.0,
435
+ "BUREAU_AMT_CREDIT_SUM_min": 110646.0,
436
+ "BUREAU_AMT_CREDIT_SUM_sum": 1549044.0,
437
+ "BUREAU_AMT_CREDIT_SUM_DEBT_max": 87473.925,
438
+ "BUREAU_AMT_CREDIT_SUM_DEBT_sum": 1224634.95,
439
+ "BUREAU_AMT_CREDIT_SUM_LIMIT_mean": 0.0,
440
+ "BUREAU_AMT_CREDIT_SUM_LIMIT_sum": 0.0,
441
+ "BUREAU_AMT_CREDIT_SUM_OVERDUE_max": 0.0,
442
+ "BUREAU_AMT_CREDIT_SUM_OVERDUE_sum": 0.0,
443
+ "BUREAU_AMT_ANNUITY_mean": 6120.0,
444
+ "BUREAU_AMT_ANNUITY_min": 6120.0,
445
+ "BUREAU_MONTHS_BALANCE_max": 0.0,
446
+ "BUREAU_CREDIT_ACTIVE_NUNIQUE": 1.0,
447
+ "BUREAU_CREDIT_CURRENCY_NUNIQUE": 1.0,
448
+ "BUREAU_CREDIT_TYPE_NUNIQUE": 1.0,
449
+ "BUREAU_STATUS_NUNIQUE": 1.0,
450
+ "BUREAU_CREDIT_ACTIVE_Bad_debt": 0.0,
451
+ "BUREAU_CREDIT_ACTIVE_Sold": 0.0,
452
+ "BUREAU_CREDIT_CURRENCY_currency_2": 0.0,
453
+ "BUREAU_CREDIT_CURRENCY_currency_3": 0.0,
454
+ "BUREAU_CREDIT_CURRENCY_currency_4": 0.0,
455
+ "BUREAU_CREDIT_TYPE_Another_type_of_loan": 0.0,
456
+ "BUREAU_CREDIT_TYPE_Car_loan": 0.0,
457
+ "BUREAU_CREDIT_TYPE_Cash_loan__non_earmarked_": 0.0,
458
+ "BUREAU_CREDIT_TYPE_Interbank_credit": 0.0,
459
+ "BUREAU_CREDIT_TYPE_Loan_for_business_development": 0.0,
460
+ "BUREAU_CREDIT_TYPE_Loan_for_purchase_of_shares__margin_lending_": 0.0,
461
+ "BUREAU_CREDIT_TYPE_Loan_for_the_purchase_of_equipment": 0.0,
462
+ "BUREAU_CREDIT_TYPE_Loan_for_working_capital_replenishment": 0.0,
463
+ "BUREAU_CREDIT_TYPE_Microloan": 0.0,
464
+ "BUREAU_CREDIT_TYPE_Mobile_operator_loan": 0.0,
465
+ "BUREAU_CREDIT_TYPE_Mortgage": 0.0,
466
+ "BUREAU_CREDIT_TYPE_Real_estate_loan": 0.0,
467
+ "BUREAU_CREDIT_TYPE_Unknown_type_of_loan": 0.0,
468
+ "BUREAU_STATUS_1": 0.0,
469
+ "BUREAU_STATUS_2": 0.0,
470
+ "BUREAU_STATUS_4": 0.0,
471
+ "BUREAU_STATUS_5": 0.0,
472
+ "PREV_AMT_ANNUITY_mean": 11570.876999999999,
473
+ "PREV_AMT_ANNUITY_min": 6750.0,
474
+ "PREV_AMT_APPLICATION_min": 0.0,
475
+ "PREV_AMT_DOWN_PAYMENT_sum": 7420.5,
476
+ "PREV_HOUR_APPR_PROCESS_START_mean": 8.7,
477
+ "PREV_NFLAG_LAST_APPL_IN_DAY_max": 1.0,
478
+ "PREV_NFLAG_LAST_APPL_IN_DAY_min": 1.0,
479
+ "PREV_RATE_DOWN_PAYMENT_max": 0.1039661288115612,
480
+ "PREV_RATE_DOWN_PAYMENT_sum": 0.1039661288115612,
481
+ "PREV_RATE_INTEREST_PRIMARY_max": None,
482
+ "PREV_RATE_INTEREST_PRIMARY_sum": 0.0,
483
+ "PREV_RATE_INTEREST_PRIVILEGED_max": None,
484
+ "PREV_SELLERPLACE_AREA_sum": 192.0,
485
+ "PREV_CNT_PAYMENT_max": 48.0,
486
+ "PREV_CNT_PAYMENT_min": 0.0,
487
+ "PREV_DAYS_FIRST_DRAWING_mean": 365243.0,
488
+ "PREV_DAYS_FIRST_DRAWING_max": 365243.0,
489
+ "PREV_DAYS_FIRST_DRAWING_sum": 730486.0,
490
+ "PREV_DAYS_FIRST_DUE_min": -547.0,
491
+ "PREV_DAYS_LAST_DUE_mean": -238.5,
492
+ "PREV_DAYS_LAST_DUE_min": -397.0,
493
+ "PREV_DAYS_TERMINATION_mean": -233.0,
494
+ "PREV_NFLAG_INSURED_ON_APPROVAL_sum": 0.0,
495
+ "PREV_NAME_PAYMENT_TYPE_NUNIQUE": 2.0,
496
+ "PREV_CODE_REJECT_REASON_NUNIQUE": 2.0,
497
+ "PREV_NAME_TYPE_SUITE_NUNIQUE": 1.0,
498
+ "PREV_NAME_CLIENT_TYPE_NUNIQUE": 2.0,
499
+ "PREV_NAME_PRODUCT_TYPE_NUNIQUE": 3.0,
500
+ "PREV_NAME_SELLER_INDUSTRY_NUNIQUE": 2.0,
501
+ "PREV_NAME_CONTRACT_TYPE_XNA": 0.0,
502
+ "PREV_WEEKDAY_APPR_PROCESS_START_FRIDAY": 0.0,
503
+ "PREV_WEEKDAY_APPR_PROCESS_START_MONDAY": 3.0,
504
+ "PREV_WEEKDAY_APPR_PROCESS_START_SATURDAY": 2.0,
505
+ "PREV_WEEKDAY_APPR_PROCESS_START_SUNDAY": 0.0,
506
+ "PREV_WEEKDAY_APPR_PROCESS_START_THURSDAY": 3.0,
507
+ "PREV_WEEKDAY_APPR_PROCESS_START_TUESDAY": 1.0,
508
+ "PREV_WEEKDAY_APPR_PROCESS_START_WEDNESDAY": 1.0,
509
+ "PREV_FLAG_LAST_APPL_PER_CONTRACT_N": 0.0,
510
+ "PREV_NAME_CASH_LOAN_PURPOSE_Building_a_house_or_an_annex": 0.0,
511
+ "PREV_NAME_CASH_LOAN_PURPOSE_Business_development": 0.0,
512
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_garage": 0.0,
513
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_holiday_home___land": 0.0,
514
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_home": 0.0,
515
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_new_car": 0.0,
516
+ "PREV_NAME_CASH_LOAN_PURPOSE_Buying_a_used_car": 0.0,
517
+ "PREV_NAME_CASH_LOAN_PURPOSE_Car_repairs": 0.0,
518
+ "PREV_NAME_CASH_LOAN_PURPOSE_Education": 0.0,
519
+ "PREV_NAME_CASH_LOAN_PURPOSE_Everyday_expenses": 0.0,
520
+ "PREV_NAME_CASH_LOAN_PURPOSE_Furniture": 0.0,
521
+ "PREV_NAME_CASH_LOAN_PURPOSE_Gasification___water_supply": 0.0,
522
+ "PREV_NAME_CASH_LOAN_PURPOSE_Hobby": 0.0,
523
+ "PREV_NAME_CASH_LOAN_PURPOSE_Journey": 0.0,
524
+ "PREV_NAME_CASH_LOAN_PURPOSE_Medicine": 0.0,
525
+ "PREV_NAME_CASH_LOAN_PURPOSE_Money_for_a_third_person": 0.0,
526
+ "PREV_NAME_CASH_LOAN_PURPOSE_Other": 1.0,
527
+ "PREV_NAME_CASH_LOAN_PURPOSE_Payments_on_other_loans": 0.0,
528
+ "PREV_NAME_CASH_LOAN_PURPOSE_Purchase_of_electronic_equipment": 0.0,
529
+ "PREV_NAME_CASH_LOAN_PURPOSE_Refusal_to_name_the_goal": 0.0,
530
+ "PREV_NAME_CASH_LOAN_PURPOSE_Repairs": 0.0,
531
+ "PREV_NAME_CASH_LOAN_PURPOSE_Urgent_needs": 0.0,
532
+ "PREV_NAME_CASH_LOAN_PURPOSE_Wedding___gift___holiday": 0.0,
533
+ "PREV_NAME_CONTRACT_STATUS_Refused": 3.0,
534
+ "PREV_NAME_CONTRACT_STATUS_Unused_offer": 0.0,
535
+ "PREV_NAME_PAYMENT_TYPE_Cashless_from_the_account_of_the_employer": 0.0,
536
+ "PREV_NAME_PAYMENT_TYPE_Non_cash_from_your_account": 0.0,
537
+ "PREV_CODE_REJECT_REASON_LIMIT": 0.0,
538
+ "PREV_CODE_REJECT_REASON_SCO": 0.0,
539
+ "PREV_CODE_REJECT_REASON_SCOFR": 0.0,
540
+ "PREV_CODE_REJECT_REASON_SYSTEM": 0.0,
541
+ "PREV_CODE_REJECT_REASON_VERIF": 0.0,
542
+ "PREV_CODE_REJECT_REASON_XNA": 0.0,
543
+ "PREV_NAME_TYPE_SUITE_Children": 0.0,
544
+ "PREV_NAME_TYPE_SUITE_Family": 0.0,
545
+ "PREV_NAME_TYPE_SUITE_Group_of_people": 0.0,
546
+ "PREV_NAME_TYPE_SUITE_Other_A": 0.0,
547
+ "PREV_NAME_TYPE_SUITE_Other_B": 0.0,
548
+ "PREV_NAME_TYPE_SUITE_Spouse__partner": 0.0,
549
+ "PREV_NAME_CLIENT_TYPE_New": 1.0,
550
+ "PREV_NAME_CLIENT_TYPE_Refreshed": 0.0,
551
+ "PREV_NAME_CLIENT_TYPE_XNA": 0.0,
552
+ "PREV_NAME_GOODS_CATEGORY_Additional_Service": 0.0,
553
+ "PREV_NAME_GOODS_CATEGORY_Animals": 0.0,
554
+ "PREV_NAME_GOODS_CATEGORY_Audio_Video": 0.0,
555
+ "PREV_NAME_GOODS_CATEGORY_Auto_Accessories": 0.0,
556
+ "PREV_NAME_GOODS_CATEGORY_Clothing_and_Accessories": 0.0,
557
+ "PREV_NAME_GOODS_CATEGORY_Computers": 0.0,
558
+ "PREV_NAME_GOODS_CATEGORY_Consumer_Electronics": 2.0,
559
+ "PREV_NAME_GOODS_CATEGORY_Direct_Sales": 0.0,
560
+ "PREV_NAME_GOODS_CATEGORY_Education": 0.0,
561
+ "PREV_NAME_GOODS_CATEGORY_Fitness": 0.0,
562
+ "PREV_NAME_GOODS_CATEGORY_Gardening": 0.0,
563
+ "PREV_NAME_GOODS_CATEGORY_Homewares": 0.0,
564
+ "PREV_NAME_GOODS_CATEGORY_Insurance": 0.0,
565
+ "PREV_NAME_GOODS_CATEGORY_Jewelry": 0.0,
566
+ "PREV_NAME_GOODS_CATEGORY_Medical_Supplies": 0.0,
567
+ "PREV_NAME_GOODS_CATEGORY_Medicine": 0.0,
568
+ "PREV_NAME_GOODS_CATEGORY_Office_Appliances": 0.0,
569
+ "PREV_NAME_GOODS_CATEGORY_Other": 0.0,
570
+ "PREV_NAME_GOODS_CATEGORY_Photo___Cinema_Equipment": 0.0,
571
+ "PREV_NAME_GOODS_CATEGORY_Sport_and_Leisure": 0.0,
572
+ "PREV_NAME_GOODS_CATEGORY_Tourism": 0.0,
573
+ "PREV_NAME_GOODS_CATEGORY_Vehicles": 0.0,
574
+ "PREV_NAME_GOODS_CATEGORY_Weapon": 0.0,
575
+ "PREV_NAME_GOODS_CATEGORY_XNA": 8.0,
576
+ "PREV_NAME_PORTFOLIO_POS": 2.0,
577
+ "PREV_NAME_PRODUCT_TYPE_walk_in": 1.0,
578
+ "PREV_CHANNEL_TYPE_AP___Cash_loan_": 0.0,
579
+ "PREV_CHANNEL_TYPE_Car_dealer": 0.0,
580
+ "PREV_CHANNEL_TYPE_Channel_of_corporate_sales": 0.0,
581
+ "PREV_CHANNEL_TYPE_Contact_center": 1.0,
582
+ "PREV_CHANNEL_TYPE_Regional___Local": 0.0,
583
+ "PREV_CHANNEL_TYPE_Stone": 2.0,
584
+ "PREV_NAME_SELLER_INDUSTRY_Auto_technology": 0.0,
585
+ "PREV_NAME_SELLER_INDUSTRY_Connectivity": 0.0,
586
+ "PREV_NAME_SELLER_INDUSTRY_Construction": 0.0,
587
+ "PREV_NAME_SELLER_INDUSTRY_Industry": 0.0,
588
+ "PREV_NAME_SELLER_INDUSTRY_Jewelry": 0.0,
589
+ "PREV_NAME_SELLER_INDUSTRY_MLM_partners": 0.0,
590
+ "PREV_NAME_SELLER_INDUSTRY_Tourism": 0.0,
591
+ "PREV_NAME_YIELD_GROUP_XNA": 6.0,
592
+ "PREV_NAME_YIELD_GROUP_low_action": 1.0,
593
+ "PREV_NAME_YIELD_GROUP_low_normal": 0.0,
594
+ "PREV_PRODUCT_COMBINATION_Card_Street": 1.0,
595
+ "PREV_PRODUCT_COMBINATION_Cash_Street__high": 0.0,
596
+ "PREV_PRODUCT_COMBINATION_Cash_Street__low": 1.0,
597
+ "PREV_PRODUCT_COMBINATION_Cash_Street__middle": 0.0,
598
+ "PREV_PRODUCT_COMBINATION_Cash_X_Sell__high": 0.0,
599
+ "PREV_PRODUCT_COMBINATION_Cash_X_Sell__middle": 1.0,
600
+ "PREV_PRODUCT_COMBINATION_POS_industry_with_interest": 0.0,
601
+ "PREV_PRODUCT_COMBINATION_POS_industry_without_interest": 0.0,
602
+ "PREV_PRODUCT_COMBINATION_POS_mobile_without_interest": 0.0,
603
+ "PREV_PRODUCT_COMBINATION_POS_other_with_interest": 0.0,
604
+ "PREV_PRODUCT_COMBINATION_POS_others_without_interest": 0.0,
605
+ "INST_PAY_NUM_INSTALMENT_VERSION_mean": 1.0588235294117647,
606
+ "INST_PAY_DAYS_ENTRY_PAYMENT_mean": -338.7647058823529,
607
+ "INST_PAY_DAYS_ENTRY_PAYMENT_min": -561.0,
608
+ "INST_PAY_AMT_PAYMENT_mean": 13291.133823529412,
609
+ "INST_PAY_AMT_PAYMENT_min": 12423.96,
610
+ "INST_PAY_AMT_PAYMENT_sum": 225949.275,
611
+ "INST_PAY_PAYMENT_DELAY_mean": -23.352941176470587,
612
+ "INST_PAY_PAYMENT_DELAY_max": -2.0,
613
+ "INST_PAY_PAYMENT_DELAY_min": -41.0,
614
+ "INST_PAY_PAYMENT_DELAY_sum": -397.0,
615
+ "INST_PAY_PAYMENT_RATIO_mean": 1.0,
616
+ "INST_PAY_PAYMENT_RATIO_min": 1.0,
617
+ "CC_MONTHS_BALANCE_mean": None,
618
+ "CC_MONTHS_BALANCE_max": None,
619
+ "CC_AMT_BALANCE_mean": None,
620
+ "CC_AMT_CREDIT_LIMIT_ACTUAL_max": None,
621
+ "CC_AMT_CREDIT_LIMIT_ACTUAL_sum": None,
622
+ "CC_AMT_DRAWINGS_ATM_CURRENT_mean": None,
623
+ "CC_AMT_DRAWINGS_ATM_CURRENT_sum": None,
624
+ "CC_AMT_DRAWINGS_CURRENT_mean": None,
625
+ "CC_AMT_DRAWINGS_CURRENT_max": None,
626
+ "CC_AMT_DRAWINGS_CURRENT_min": None,
627
+ "CC_AMT_DRAWINGS_POS_CURRENT_min": None,
628
+ "CC_AMT_INST_MIN_REGULARITY_min": None,
629
+ "CC_AMT_PAYMENT_CURRENT_min": None,
630
+ "CC_AMT_PAYMENT_TOTAL_CURRENT_max": None,
631
+ "CC_CNT_DRAWINGS_ATM_CURRENT_mean": None,
632
+ "CC_CNT_DRAWINGS_ATM_CURRENT_min": None,
633
+ "CC_CNT_DRAWINGS_ATM_CURRENT_sum": None,
634
+ "CC_CNT_DRAWINGS_CURRENT_max": None,
635
+ "CC_CNT_DRAWINGS_CURRENT_min": None,
636
+ "CC_CNT_DRAWINGS_OTHER_CURRENT_mean": None,
637
+ "CC_CNT_DRAWINGS_OTHER_CURRENT_min": None,
638
+ "CC_SK_DPD_max": None,
639
+ "CC_SK_DPD_DEF_max": None,
640
+ "CC_UTILIZATION_mean": None,
641
+ "CC_UTILIZATION_max": None,
642
+ "CC_UTILIZATION_min": None,
643
+ "CC_PAYMENT_RATIO_max": None,
644
+ "CC_PAYMENT_RATIO_min": None,
645
+ "CC_NAME_CONTRACT_STATUS_NUNIQUE": None,
646
+ "CC_NAME_CONTRACT_STATUS_Approved": None,
647
+ "CC_NAME_CONTRACT_STATUS_Completed": None,
648
+ "CC_NAME_CONTRACT_STATUS_Refused": None,
649
+ "CC_NAME_CONTRACT_STATUS_Sent_proposal": None,
650
+ "CC_NAME_CONTRACT_STATUS_Signed": None,
651
+ "POS_CNT_INSTALMENT_min": 6.0,
652
+ "POS_CNT_INSTALMENT_FUTURE_min": 0.0,
653
+ "POS_SK_DPD_mean": 0.0,
654
+ "POS_SK_DPD_min": 0.0,
655
+ "POS_SK_DPD_DEF_max": 0.0,
656
+ "POS_SK_DPD_DEF_min": 0.0,
657
+ "POS_NAME_CONTRACT_STATUS_NUNIQUE": 3.0,
658
+ "POS_NAME_CONTRACT_STATUS_Amortized_debt": 0.0,
659
+ "POS_NAME_CONTRACT_STATUS_Approved": 0.0,
660
+ "POS_NAME_CONTRACT_STATUS_Canceled": 0.0,
661
+ "POS_NAME_CONTRACT_STATUS_Completed": 2.0,
662
+ "POS_NAME_CONTRACT_STATUS_Returned_to_the_store": 0.0,
663
+ "POS_NAME_CONTRACT_STATUS_Signed": 1.0,
664
+ "POS_NAME_CONTRACT_STATUS_XNA": 0.0
665
+ }
666
+ }
667
+ }
Dockerfile ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # force rebuild
2
+ FROM python:3.11.9-slim
3
+
4
+ ENV PYTHONDONTWRITEBYTECODE=1
5
+ ENV PYTHONUNBUFFERED=1
6
+
7
+ WORKDIR /App
8
+
9
+ COPY requirements.txt .
10
+ RUN pip install --no-cache-dir -r requirements.txt
11
+
12
+ # Copie des fichers nécessaires
13
+ COPY App ./App
14
+ COPY best_threshold.joblib .
15
+
16
+ EXPOSE 8000
17
+
18
+ CMD ["uvicorn", "App.main:app", "--host", "0.0.0.0", "--port", "8000"]
Images/Approche_mlops.png ADDED
Images/CM_modele_retenu.png ADDED

Git LFS Details

  • SHA256: c9285930945c3f4792aede34b3493d5cc96546d662973e45c55aaa43f61bf879
  • Pointer size: 131 Bytes
  • Size of remote file: 293 kB
Images/Pred_finales.png ADDED
Images/SHAP_global.png ADDED

Git LFS Details

  • SHA256: 7a412df4430ad15fa11ade315de15e3c47a5301632163fe53d053cb77825e719
  • Pointer size: 132 Bytes
  • Size of remote file: 1.28 MB
Images/SHAP_local.png ADDED

Git LFS Details

  • SHA256: 6f818dea3713a79e9f1f5d6b0bf9a68daedbbdc0b46f91317c68e24b63d949d5
  • Pointer size: 131 Bytes
  • Size of remote file: 988 kB
Images/Schemas_tables.png ADDED
Images/steps_ops_modeles.png ADDED
Notesbooks/Notebook_Cleaning-Viz_022026.ipynb ADDED
The diff for this file is too large to render. See raw diff
 
Notesbooks/Notebook_Data-Preparation_022026.ipynb ADDED
The diff for this file is too large to render. See raw diff
 
Notesbooks/Notebook_Features-Imp-SHAP_022026.ipynb ADDED
The diff for this file is too large to render. See raw diff
 
Notesbooks/Notebook_Model-Training-MLFlow_022026.ipynb ADDED
The diff for this file is too large to render. See raw diff
 
README.md CHANGED
@@ -1,10 +1,212 @@
1
- ---
2
- title: Home Credit API
3
- emoji: 👀
4
- colorFrom: yellow
5
- colorTo: blue
6
- sdk: docker
7
- pinned: false
8
- ---
9
-
10
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Classification Risque Crédit - Pipeline MLOps
2
+
3
+ ## Contexte
4
+ L’entreprise souhaite mettre en place un système automatisé de scoring crédit capable :
5
+ - d’estimer la probabilité de défaut d’un client,
6
+ - de classer automatiquement une demande en crédit accordé ou crédit refusé,
7
+ - tout en garantissant transparence et industrialisation du cycle de vie du modèle.
8
+
9
+ Ce projet s’inscrit dans une démarche MLOps complète, du tracking d’expérimentations jusqu’au serving du modèle.
10
+
11
+ ## Objectifs du projet
12
+
13
+ Ce projet a pour objectif principal de **construire et optimiser un modèle de scoring prédictif du risque de défaut**.
14
+
15
+ Autrement dit, il faudra:
16
+ - Analyser les variables explicatives:
17
+ - Feature importance globale
18
+ - Explicabilité locale (niveau client)
19
+
20
+ - Implémenter une démarche MLOps complète:
21
+ - Tracking des expérimentations
22
+ - Model Registry
23
+ - Serving
24
+
25
+ - Optimiser un score métier prenant en compte:
26
+ - Le déséquilibre de classes
27
+ - Le coût asymétrique des erreurs (FN = 10 × FP)
28
+
29
+ - Comparer plusieurs modèles via Cross-Validation et GridSearch.
30
+
31
+ ## Sources de données
32
+
33
+ ![Lien entre les tables](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/Schemas_tables.png)
34
+
35
+ Nous avons reçu des données provenant de sources multiples:
36
+ - Socio‑démographiques
37
+ - Historiques de crédit
38
+ - Comportements de paiement
39
+ - Interactions avec d’autres institutions
40
+
41
+ ## Nettoyage structurel des données
42
+
43
+ Avant toute modélisation, un nettoyage approfondi a été réalisé pour garantir la qualité des données :
44
+
45
+ **1. Harmonisation des types**
46
+ - Conversion des dates en format datetime
47
+ - Uniformisation des variables catégorielles
48
+ - Correction des types numériques (float/int)
49
+
50
+ **2. Préparation des valeurs manquantes & outliers**
51
+
52
+ L’objectif ici n’est pas de *“nettoyer à la main”*, mais de préparer les données pour qu’elles soient
53
+ traitées de manière systématique dans le pipeline MLOps, sans logique ad hoc.
54
+
55
+ **3. Nettoyage structurel**
56
+ - Harmonisation des noms de varibales et modifications des formats
57
+ - Suppression des colonnes constantes ou quasi‑constantes
58
+ - Suppression des identifiants non pertinents
59
+ - Vérification des doublons
60
+ - Fusion des tables <> obtenir un seul historique complet des comportements des clients.
61
+
62
+ ## Feature Engineering
63
+
64
+ Pour chaque client, des agrégations (*min, max, mean, count, sum*) ont été calculées sur les tables.
65
+
66
+ Concernant les colonnes crées, nous avons choisi d'intéger des varialbles majoritairement liées au comportement réel de remboursement du crédit:
67
+ - **PAYMENT_DELAY**: retard réel vs prévu
68
+ - **LATE_PAYMENT**: indicateur de retard
69
+ - **PARTIAL_PAYMENT**: paiement incomplet
70
+ - **EARLY_PAYMENT**: paiement anticipé
71
+ - **PAYMENT_RATIO**: montant payé / montant
72
+ - **UTILIZATION**: utilisation du crédit
73
+ PAYMENT_RATIO : ratio paiement / retraits
74
+
75
+
76
+ ## Analyse exploratoire — Conclusions clés
77
+ L’analyse exploratoire a permis d’identifier plusieurs signaux forts sur le profil des *mauvais payeurs*:
78
+
79
+ **1. Variables socio‑démographiques**
80
+ - Revenus faibles présentent un taux de défaut plus élevé
81
+ - Les célibataires et divorcés sont plus à risque
82
+ - Les niveaux d’éducation inférieurs sont associés à un risque accru
83
+
84
+ **2. Historique de crédit**
85
+ - Les clients ayant déjà eu des retards plus risqués
86
+ - La capacité de remboursement du client est un des indicateurs forts
87
+ - Un ratio payé/dû traduisant des paiements irréguliers.
88
+
89
+ **3. Comportement de paiement**
90
+ - Les refus antérieurs sont très prédictifs
91
+ - Les nouveaux clients sont plus incertains
92
+ - Les retards successifs sont très discriminants
93
+
94
+ ## Approche Machine Learning Operations (**MLOps**)
95
+
96
+ Le projet suit une démarche MLOps complète :
97
+
98
+ **1. Tracking des expérimentations (MLflow)**
99
+ - Suivi des hyperparamètres
100
+ - Suivi des métriques (AUC, recall, coût métier)
101
+ - Versioning des modèles
102
+ - Export des artefacts (ROC, CM, SHAP)
103
+
104
+ **2. Organisation des experiments**
105
+ - Baseline
106
+ - Feature engineering
107
+ - Traitement des outliers
108
+ - Optimisation des hyperparamètres
109
+ - Sélection du meilleur modèle via coût métier
110
+
111
+ **3. Reproductibilité**
112
+ - Environnement Conda versionné
113
+ - Scripts d’entraînement reproductibles
114
+ - Artefacts exportés dans *artefacts/*
115
+
116
+ Le projet repose donc sur une approche MLOps qui répond à ce schéma:
117
+
118
+ ![MLOps](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/Approche_mlops.png)
119
+
120
+ **Préprocessings initiaux**
121
+ - un ***modèle de base***, puis
122
+ - un ***modèle optimisé***, via une recherche d’hyperparamètres (RandomizedSearchCV)
123
+ pour chaque preprocessing constituant une branche indépendante & testée de manière équitable.
124
+
125
+ Les processings initiaux sont:
126
+ - **Missing Flag**: conserver les valeurs manquantes en les rendant explicites
127
+ - **Imputation**: remplacer les valeurs manquantes par des statistiques adaptées
128
+ - **Drop70**: supprimer les variables trop incomplètes pour réduire le bruit
129
+
130
+ **Nettoyage des outliers**
131
+ Etape transversale pour mesurer l’impact réel du nettoyage des valeurs aberrantes sur chaque preprocessing.
132
+
133
+ **Transformation des colonnes**
134
+ - **StandardScaler** sur les variables numériques,
135
+ - **OneHotEncoder** sur les variables catégorielles nominales,
136
+ - **TargetEncoder** sur les variables catégorielles à forte cardinalité.
137
+
138
+ **Entraînement**
139
+ Chaque pipeline complet est évalué selon trois axes :
140
+ - **Coût métier**: critère principal, basé sur le coût des faux positifs et faux négatifs
141
+ - **Seuil optimal**: déterminé pour minimiser le coût global.
142
+ - **Métriques classiques**: AUC, recall, precision, F1, etc.
143
+
144
+ Le modèle final est celui qui obtient le **coût métier minimal**, toutes expériences confondues.
145
+
146
+ ## Sélection du meilleur modèle
147
+
148
+ Plusieurs modèles ont été testés allant du plus simple (*régression logistique*) au plus avancé (*XGBoost*). Le *Dummy* sert de référence, la *régression logistique* est interprétable mais limitée, le *Random Forest* capturent les relations complexes, et *XGBoost* offre le meilleur compromis entre performance et robustesse.
149
+
150
+ ![steps](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/steps_ops_modeles.png)
151
+
152
+ **Critère principal: coût métier = ```10 x FN + 1 x FP```**
153
+
154
+ **Modèle retenu: XGBoost optimisé**
155
+
156
+ | Résultats | Valeurs |
157
+ |-----------|-----------|
158
+ | AUC | **0.78** |
159
+ | Recall mauvais payeurs | **67%** |
160
+ | Coût métier minimal | **30 629 €** |
161
+ | Seuil optimal | **0.5** |
162
+ | | |
163
+
164
+ Le modèle permet:
165
+
166
+ - d’identifier 3319 mauvais payeurs sur 4965
167
+
168
+ - de réduire les faux négatifs (erreurs les plus coûteuses)
169
+
170
+ - de maintenir un bon équilibre entre détection du risque et acceptation des bons clients.
171
+
172
+ ![CM](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/CM_modele_retenu.png)
173
+
174
+ ## Feature Importance & Explicabilité (**SHAP**)
175
+
176
+ **Variables les plus influentes**
177
+
178
+ - Scores externes de solvabilité
179
+ - Historique de crédit
180
+ - Retards de paiement
181
+ - Refus antérieurs
182
+ - Ratios de paiement (installments)
183
+
184
+ ![Global](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/SHAP_global.png)
185
+
186
+ **Explicabilité locale**
187
+ Pour chaque client, les valeurs SHAP permettent d’expliquer:
188
+ - Pourquoi le score augmente
189
+ - Pourquoi le score diminue
190
+ - Quelles variables influencent le plus la décision
191
+
192
+ ![Local](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/SHAP_local.png)
193
+
194
+ ## Prédictions finales
195
+
196
+ Le modèle retenu (*XGBoost optimisé*) a été appliqué à un jeu de données inédit, non utilisé lors de l’entraînement.
197
+
198
+ L’objectif est d’évaluer son comportement en conditions réelles et de produire une décision de crédit automatisée.
199
+
200
+ ![Prédictions finales](https://raw.githubusercontent.com/Diaure/Classification-Risque-Credit-Pipeline-MLOps-/master/Images/Pred_finales.png)
201
+
202
+ Cette forte proportion de refus est cohérente avec:
203
+
204
+ - le seuil optimal choisi pour maximiser le rappel des mauvais payeurs,
205
+ - la distribution naturellement risquée du dataset,
206
+ - la stratégie volontairement conservatrice du modèle (réduire les FN).
207
+
208
+ ## CI/CD et Déploiement
209
+
210
+ Ce projet met en œuvre une approche CI/CD complète, séparant:
211
+ - l’intégration continue (**CI**): garantir la qualité du code
212
+ - le déploiement continu (**CD**): rendre l’API accessible publiquement
best_threshold.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fac184ccb7c4dfd54f294c0ddc5e73806500dc64a63937144609988937252eb7
3
+ size 116
deploy_to_hf.py ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from huggingface_hub import HfApi
2
+ import os
3
+
4
+ api = HfApi()
5
+
6
+ api.upload_folder(
7
+ folder_path=".",
8
+ repo_id="Diaure/Home_Credit_API",
9
+ repo_type="space",
10
+ token=os.environ["HF_TOKEN"],
11
+ commit_message="CD: update from GitHub main",
12
+ ignore_patterns=[".git/*", ".github/*", "tests/*", "__pycache__/*",
13
+ "*.pyc", ".venv/*", "venv/*", "BestModel/*" ])
14
+ print(">>> CD HuggingFace: script lancé")
environment.yml ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: mlops
2
+ channels:
3
+ - defaults
4
+ - https://repo.anaconda.com/pkgs/main
5
+ - https://repo.anaconda.com/pkgs/r
6
+ - https://repo.anaconda.com/pkgs/msys2
7
+ dependencies:
8
+ - blas=1.0
9
+ - bzip2=1.0.8
10
+ - ca-certificates=2025.12.2
11
+ - intel-openmp=2025.0.0
12
+ - libffi=3.4.4
13
+ - libhwloc=2.12.1
14
+ - libiconv=1.16
15
+ - libxml2=2.13.9
16
+ - libzlib=1.3.1
17
+ - mkl=2025.0.0
18
+ - mkl-service=2.5.2
19
+ - mkl_fft=2.1.1
20
+ - mkl_random=1.3.0
21
+ - openssl=3.0.19
22
+ - packaging=25.0
23
+ - pip=25.3
24
+ - python=3.11.9
25
+ - python-dateutil=2.9.0post0
26
+ - python-tzdata=2025.3
27
+ - setuptools=80.10.2
28
+ - six=1.17.0
29
+ - sqlite=3.51.1
30
+ - tbb=2022.3.0
31
+ - tbb-devel=2022.3.0
32
+ - tk=8.6.15
33
+ - tzdata=2025c
34
+ - ucrt=10.0.22621.0
35
+ - vc=14.3
36
+ - vc14_runtime=14.44.35208
37
+ - vs2015_runtime=14.44.35208
38
+ - wheel=0.46.3
39
+ - xz=5.6.4
40
+ - zlib=1.3.1
41
+ - pip:
42
+ - alembic==1.18.4
43
+ - annotated-doc==0.0.4
44
+ - annotated-types==0.7.0
45
+ - anyio==4.12.1
46
+ - asttokens==3.0.1
47
+ - backports-tarfile==1.2.0
48
+ - blinker==1.9.0
49
+ - build==1.4.0
50
+ - cachecontrol==0.14.4
51
+ - cachetools==6.2.6
52
+ - catboost==1.2.7
53
+ - category-encoders==2.6.3
54
+ - certifi==2026.1.4
55
+ - cffi==2.0.0
56
+ - charset-normalizer==3.4.4
57
+ - cleo==2.1.0
58
+ - click==8.3.1
59
+ - cloudpickle==3.1.2
60
+ - colorama==0.4.6
61
+ - colorlog==6.10.1
62
+ - comm==0.2.3
63
+ - contourpy==1.3.3
64
+ - coverage==7.13.4
65
+ - crashtest==0.4.1
66
+ - cryptography==46.0.5
67
+ - cycler==0.12.1
68
+ - databricks-sdk==0.87.0
69
+ - debugpy==1.8.20
70
+ - decorator==5.2.1
71
+ - distlib==0.4.0
72
+ - docker==7.1.0
73
+ - dulwich==1.0.0
74
+ - executing==2.2.1
75
+ - fastapi==0.115.0
76
+ - fastjsonschema==2.21.2
77
+ - filelock==3.24.0
78
+ - findpython==0.7.1
79
+ - flask==3.1.2
80
+ - flask-cors==6.0.2
81
+ - fonttools==4.61.1
82
+ - fsspec==2026.2.0
83
+ - gitdb==4.0.12
84
+ - gitpython==3.1.46
85
+ - google-auth==2.48.0
86
+ - graphene==3.4.3
87
+ - graphql-core==3.2.7
88
+ - graphql-relay==3.2.0
89
+ - greenlet==3.3.1
90
+ - h11==0.16.0
91
+ - hf-xet==1.2.0
92
+ - httpcore==1.0.9
93
+ - httpx==0.28.1
94
+ - huey==2.6.0
95
+ - huggingface-hub==1.3.1
96
+ - idna==3.11
97
+ - importlib-metadata==8.7.1
98
+ - iniconfig==2.3.0
99
+ - installer==0.7.0
100
+ - ipykernel==6.31.0
101
+ - ipython==9.10.0
102
+ - ipython-pygments-lexers==1.1.1
103
+ - itsdangerous==2.2.0
104
+ - jaraco-classes==3.4.0
105
+ - jaraco-context==6.1.0
106
+ - jaraco-functools==4.4.0
107
+ - jedi==0.19.2
108
+ - jinja2==3.1.6
109
+ - joblib==1.5.3
110
+ - jupyter-client==8.8.0
111
+ - jupyter-core==5.9.1
112
+ - keyring==25.7.0
113
+ - kiwisolver==1.4.9
114
+ - llvmlite==0.42.0
115
+ - mako==1.3.10
116
+ - markdown-it-py==4.0.0
117
+ - markupsafe==3.0.3
118
+ - matplotlib==3.8.4
119
+ - matplotlib-inline==0.2.1
120
+ - mdurl==0.1.2
121
+ - mlflow==3.9.0
122
+ - mlflow-skinny==3.9.0
123
+ - mlflow-tracing==3.9.0
124
+ - more-itertools==10.8.0
125
+ - msgpack==1.1.2
126
+ - narwhals==2.16.0
127
+ - nest-asyncio==1.6.0
128
+ - numba==0.59.1
129
+ - numpy==1.26.4
130
+ - opentelemetry-api==1.39.1
131
+ - opentelemetry-proto==1.39.1
132
+ - opentelemetry-sdk==1.39.1
133
+ - opentelemetry-semantic-conventions==0.60b1
134
+ - optuna==3.6.1
135
+ - pandas==2.2.3
136
+ - parso==0.8.5
137
+ - patsy==1.0.2
138
+ - pbs-installer==2026.2.11
139
+ - pillow==12.1.0
140
+ - pkginfo==1.12.1.2
141
+ - platformdirs==4.5.1
142
+ - plotly==6.5.2
143
+ - pluggy==1.6.0
144
+ - poetry==2.3.2
145
+ - poetry-core==2.3.1
146
+ - prettytable==3.17.0
147
+ - prompt-toolkit==3.0.52
148
+ - protobuf==6.33.5
149
+ - psutil==7.2.2
150
+ - psycopg2-binary==2.9.11
151
+ - pure-eval==0.2.3
152
+ - pyarrow==22.0.0
153
+ - pyasn1==0.6.2
154
+ - pyasn1-modules==0.4.2
155
+ - pycparser==3.0
156
+ - pydantic==2.12.5
157
+ - pydantic-core==2.41.5
158
+ - pygments==2.19.2
159
+ - pyparsing==3.3.2
160
+ - pyproject-hooks==1.2.0
161
+ - pytest==9.0.2
162
+ - pytest-cov==7.0.0
163
+ - python-dotenv==1.2.1
164
+ - python-graphviz==0.21
165
+ - pytz==2025.2
166
+ - pywin32==311
167
+ - pywin32-ctypes==0.2.3
168
+ - pyyaml==6.0.3
169
+ - pyzmq==27.1.0
170
+ - rapidfuzz==3.14.3
171
+ - requests==2.32.5
172
+ - requests-toolbelt==1.0.0
173
+ - rich==14.3.2
174
+ - rsa==4.9.1
175
+ - scikit-learn==1.4.2
176
+ - scipy==1.12.0
177
+ - seaborn==0.13.2
178
+ - shap==0.49.1
179
+ - shellingham==1.5.4
180
+ - skops==0.13.0
181
+ - slicer==0.0.8
182
+ - smmap==5.0.2
183
+ - sqlalchemy==2.0.46
184
+ - sqlparse==0.5.5
185
+ - stack-data==0.6.3
186
+ - starlette==0.38.6
187
+ - statsmodels==0.14.6
188
+ - threadpoolctl==3.6.0
189
+ - tomlkit==0.14.0
190
+ - tornado==6.5.4
191
+ - tqdm==4.67.3
192
+ - traitlets==5.14.3
193
+ - trove-classifiers==2026.1.14.14
194
+ - typer==0.23.1
195
+ - typer-slim==0.23.1
196
+ - typing-extensions==4.15.0
197
+ - typing-inspection==0.4.2
198
+ - urllib3==2.6.3
199
+ - uvicorn==0.30.1
200
+ - virtualenv==20.36.1
201
+ - waitress==3.0.2
202
+ - wcwidth==0.6.0
203
+ - werkzeug==3.1.5
204
+ - xgboost==2.0.3
205
+ - zipp==3.23.0
206
+ - zstandard==0.25.0
207
+ prefix: C:\Users\Lenovo\anaconda3\envs\mlops
pipeline_complet.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:323bdb87282d8efdebb0c9f51c1ae7eba2aedb014281fae92ae448ee7fcbb0d6
3
+ size 593634
poetry.lock ADDED
The diff for this file is too large to render. See raw diff
 
pyproject.toml ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [project]
2
+ name = "projet6"
3
+ version = "0.1.0"
4
+ description = ""
5
+ authors = [
6
+ {name = "Aurélie GABU",email = "aurelie.gabu@gmail.com"}
7
+ ]
8
+ requires-python = ">=3.11,<3.12"
9
+ dependencies = [
10
+ "pandas ==2.2.3",
11
+ "numpy ==1.26.4",
12
+ "scikit-learn ==1.4.2",
13
+ "matplotlib ==3.8.4",
14
+ "seaborn ==0.13.2",
15
+ "scipy ==1.12.0",
16
+ "shap ==0.49.1",
17
+ "xgboost ==2.0.3",
18
+ "optuna ==3.6.1",
19
+ "category-encoders ==2.6.3",
20
+ "catboost ==1.2.7",
21
+ "numba ==0.59.1",
22
+ "llvmlite ==0.42.0",
23
+ "ipykernel>=6.25,<7.0",
24
+ "huggingface-hub ==1.3.1",
25
+ "fastapi ==0.115.0",
26
+ "uvicorn ==0.30.1",
27
+ "python-dotenv ==1.2.1",
28
+ "psycopg2-binary ==2.9.11",
29
+ "pytest-cov ==7.0.0"
30
+ ]
31
+
32
+
33
+ [build-system]
34
+ requires = ["poetry-core>=2.0.0,<3.0.0"]
35
+ build-backend = "poetry.core.masonry.api"
requirements.txt ADDED
@@ -0,0 +1,168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ alembic==1.18.4
2
+ annotated-doc==0.0.4
3
+ annotated-types==0.7.0
4
+ anyio==4.12.1
5
+ asttokens==3.0.1
6
+ backports.tarfile==1.2.0
7
+ blinker==1.9.0
8
+ build==1.4.0
9
+ CacheControl==0.14.4
10
+ cachetools==6.2.6
11
+ catboost==1.2.7
12
+ category-encoders==2.6.3
13
+ certifi==2026.1.4
14
+ cffi==2.0.0
15
+ charset-normalizer==3.4.4
16
+ cleo==2.1.0
17
+ click==8.3.1
18
+ cloudpickle==3.1.2
19
+ colorama==0.4.6
20
+ colorlog==6.10.1
21
+ comm==0.2.3
22
+ contourpy==1.3.3
23
+ coverage==7.13.4
24
+ crashtest==0.4.1
25
+ cryptography==46.0.5
26
+ cycler==0.12.1
27
+ databricks-sdk==0.87.0
28
+ debugpy==1.8.20
29
+ decorator==5.2.1
30
+ distlib==0.4.0
31
+ docker==7.1.0
32
+ dulwich==1.0.0
33
+ executing==2.2.1
34
+ fastapi==0.115.0
35
+ fastjsonschema==2.21.2
36
+ filelock==3.24.0
37
+ findpython==0.7.1
38
+ Flask==3.1.2
39
+ flask-cors==6.0.2
40
+ fonttools==4.61.1
41
+ fsspec==2026.2.0
42
+ gitdb==4.0.12
43
+ GitPython==3.1.46
44
+ google-auth==2.48.0
45
+ graphene==3.4.3
46
+ graphql-core==3.2.7
47
+ graphql-relay==3.2.0
48
+ graphviz==0.21
49
+ greenlet==3.3.1
50
+ h11==0.16.0
51
+ hf-xet==1.2.0
52
+ httpcore==1.0.9
53
+ httpx==0.28.1
54
+ huey==2.6.0
55
+ huggingface_hub==1.3.1
56
+ idna==3.11
57
+ importlib_metadata==8.7.1
58
+ iniconfig==2.3.0
59
+ installer==0.7.0
60
+ ipykernel==6.31.0
61
+ ipython==9.10.0
62
+ ipython_pygments_lexers==1.1.1
63
+ itsdangerous==2.2.0
64
+ jaraco.classes==3.4.0
65
+ jaraco.context==6.1.0
66
+ jaraco.functools==4.4.0
67
+ jedi==0.19.2
68
+ Jinja2==3.1.6
69
+ joblib==1.5.3
70
+ jupyter_client==8.8.0
71
+ jupyter_core==5.9.1
72
+ keyring==25.7.0
73
+ kiwisolver==1.4.9
74
+ llvmlite==0.42.0
75
+ Mako==1.3.10
76
+ markdown-it-py==4.0.0
77
+ MarkupSafe==3.0.3
78
+ matplotlib==3.8.4
79
+ matplotlib-inline==0.2.1
80
+ mdurl==0.1.2
81
+ mkl-service==2.5.2
82
+ mlflow==3.9.0
83
+ mlflow-skinny==3.9.0
84
+ mlflow-tracing==3.9.0
85
+ more-itertools==10.8.0
86
+ msgpack==1.1.2
87
+ narwhals==2.16.0
88
+ nest-asyncio==1.6.0
89
+ numba==0.59.1
90
+ numpy==1.26.4
91
+ opentelemetry-api==1.39.1
92
+ opentelemetry-proto==1.39.1
93
+ opentelemetry-sdk==1.39.1
94
+ opentelemetry-semantic-conventions==0.60b1
95
+ optuna==3.6.1
96
+ packaging==25.0
97
+ pandas==2.2.3
98
+ parso==0.8.5
99
+ patsy==1.0.2
100
+ pbs-installer==2026.2.11
101
+ pillow==12.1.0
102
+ pkginfo==1.12.1.2
103
+ platformdirs==4.5.1
104
+ plotly==6.5.2
105
+ pluggy==1.6.0
106
+ poetry==2.3.2
107
+ poetry-core==2.3.1
108
+ prettytable==3.17.0
109
+ prompt_toolkit==3.0.52
110
+ protobuf==6.33.5
111
+ psutil==7.2.2
112
+ psycopg2-binary==2.9.11
113
+ pure_eval==0.2.3
114
+ pyarrow==22.0.0
115
+ pyasn1==0.6.2
116
+ pyasn1_modules==0.4.2
117
+ pycparser==3.0
118
+ pydantic==2.12.5
119
+ pydantic_core==2.41.5
120
+ Pygments==2.19.2
121
+ pyparsing==3.3.2
122
+ pyproject_hooks==1.2.0
123
+ pytest==9.0.2
124
+ pytest-cov==7.0.0
125
+ python-dateutil==2.9.0.post0
126
+ python-dotenv==1.2.1
127
+ pytz==2025.2
128
+ PyYAML==6.0.3
129
+ pyzmq==27.1.0
130
+ RapidFuzz==3.14.3
131
+ requests==2.32.5
132
+ requests-toolbelt==1.0.0
133
+ rich==14.3.2
134
+ rsa==4.9.1
135
+ scikit-learn==1.4.2
136
+ scipy==1.12.0
137
+ seaborn==0.13.2
138
+ shap==0.49.1
139
+ shellingham==1.5.4
140
+ six==1.17.0
141
+ skops==0.13.0
142
+ slicer==0.0.8
143
+ smmap==5.0.2
144
+ SQLAlchemy==2.0.46
145
+ sqlparse==0.5.5
146
+ stack-data==0.6.3
147
+ starlette==0.38.6
148
+ statsmodels==0.14.6
149
+ threadpoolctl==3.6.0
150
+ tomlkit==0.14.0
151
+ tornado==6.5.4
152
+ tqdm==4.67.3
153
+ traitlets==5.14.3
154
+ trove-classifiers==2026.1.14.14
155
+ typer==0.23.1
156
+ typer-slim==0.23.1
157
+ typing-inspection==0.4.2
158
+ typing_extensions==4.15.0
159
+ tzdata==2025.3
160
+ urllib3==2.6.3
161
+ uvicorn==0.30.1
162
+ virtualenv==20.36.1
163
+ waitress==3.0.2
164
+ wcwidth==0.6.0
165
+ Werkzeug==3.1.5
166
+ xgboost==2.0.3
167
+ zipp==3.23.0
168
+ zstandard==0.25.0